Compare commits

..
Author SHA1 Message Date
Mike Reeves 6d7be47a69 Trim Telegraf collection on Postgres-only grids and add Postgres support to Elasticsearch growth scripts 2026-11-17 14:17:27 -05:00
Josh Patterson e51706acad Merge pull request #16399 from Security-Onion-Solutions/fix/auto-apply-push-result-unknown
FIX: queue auto state apply pushes
2026-11-17 10:53:39 -05:00
Jorge Reyes 9612b15fca Merge pull request #16400 from Security-Onion-Solutions/reyesj2-patch-2
remove unused localrules config
2026-11-16 14:58:56 -06:00
reyesj2 0f5170e17e soup missing closing } post merge resolution; 2026-11-16 14:19:22 -06:00
reyesj2 cac027d438 fix merge 2026-11-16 13:52:05 -06:00
Jorge Reyes dea6932342 Merge branch '3/dev' into reyesj2-patch-2 2026-11-16 13:50:14 -06:00
Mike Reeves 043b2185d6 Merge pull request #16398 from Security-Onion-Solutions/foundry/telegraf-partition-11tqi
Postgres: Telegraf partition maintenance over the socket, with grid-test fixes
2026-11-16 14:47:37 -05:00
reyesj2 e7bb00d19d move /nsm/rules ownership into soc/config.sls. nginx no longer serves /nsm/rules, but so-soc still mounts it read-write 2026-11-16 13:47:25 -06:00
reyesj2 8c3611ac72 only rewrite pillar if it has localrules reference(s) 2026-11-16 13:46:10 -06:00
reyesj2 5558c54632 so-yaml: treat non-mapping parents as missing keys
getKeyValue and removeKey now return early when a parent of the key is
null or not a mapping, so get, remove and purgelistitem report the key as
missing instead of crashing on a hand-edited pillar such as
`firewall: null`.
2026-11-16 13:44:58 -06:00
Josh Patterson 9045aff88d FIX: correct the batch default in the pillar_push_map.yaml comment
The comment gave the orch default batch as 25%. salt/salt/defaults.yaml
sets salt:auto_apply:batch to 10%.
2026-11-16 14:19:51 -05:00
Josh Patterson cc9dd0a101 FIX: drop malformed push_dispatched records instead of failing the pass
so-push-drainer computed a record's age before the per-record
try/except. A record whose dispatched_at or checked_at was not a number
raised TypeError there, so every pass exited before the drain and
auto-apply stopped. The traceback went to the stderr of a schedule with
return_job: False, so nothing reached so-push-drainer.log.

Only the drainer writes these records, so this takes a hand-edited file,
but result tracking is not supposed to be able to block the drain. Log a
warning and remove the record, as is already done for unreadable JSON
and a missing jid.

Also reword the so-log-check exclusion comment, which said the queued
state run "still completes" where the drainer only claims it may have,
and note at the drainer's warning that so-log-check matches its text.
2026-11-16 14:19:51 -05:00
Mike Reeves 52f9fe3092 Remove blank line in soup script 2026-11-16 13:52:14 -05:00
Mike Reeves f492e87f20 Add check for post-upgrade version file 2026-11-16 12:12:11 -05:00
Mike Reeves 01518ec25e Fix formatting issues in soup script 2026-11-16 11:14:06 -05:00
Jorge Reyes 5f3c1c84a4 Merge pull request #16396 from Security-Onion-Solutions/reyesj2-patch-3
elastic artifact repo only needs /beats/ dir
2026-11-16 09:22:23 -06:00
Josh Patterson 70053e362c FIX: exclude the drainer's push-result-unknown warning from so-log-check
so-push-drainer logs "push result unknown" as a WARNING when salt-master
restarts while a pushed state run is queued: the orchestration's next
publish fails with AuthenticationError, but the minion still runs the
queued state. so-log-check matches any line containing "error", so the
quoted exception still failed the check.

Reproduced on a distributed 3.5.0 grid: a sensoroni push queued behind a
150s state run on the manager, then salt-master was restarted. The
orchestration (which runs under salt-minion.service, so it survives the
restart) raised AuthenticationError 9s later, the drainer logged the
WARNING, and the queued state run completed with 0 failed states.
so-log-check --unknown flagged the line.

The pattern requires AuthenticationError so an unknown result from any
other raised exception, e.g. a bug in a state module, is still reported.
2026-11-16 10:20:05 -05:00
Mike Reeves b01f01da82 Fix formatting issues in soup script 2026-11-13 16:07:55 -05:00
Mike Reeves 03317ba5e4 Postgres: Telegraf partition maintenance over the socket, with grid-test fixes 2026-11-12 20:42:29 -05:00
Josh Patterson cc970892ed FIX: keep the app's target when pushing a per-node setting
For a setting saved for a single node, push_pillar replaced the app's
target from pillar_push_map.yaml with the node id. The app's state was
then applied to that node even if the node does not run the app.

Seen in the lab: the SOC config test saves
elasticsearch.config.cluster.routing.allocation.disk.threshold_enabled
for the first node SOC lists, and SOC lists nodes in random order. When
that node was a sensor, the elasticsearch state was applied to it and
failed to render (so_roles has no so-sensor entry), and so-push-drainer
logged "push failed".

Narrow the app's target to the node instead of replacing it:
"L@<node> and ( <app target> )". A node outside the app's target then
matches nothing, and so-push-drainer skips the action.
2026-11-11 10:07:08 -05:00
reyesj2 4e0558d2d3 remove unused localrules config 2026-11-10 17:30:01 -06:00
Josh Patterson 22a4acbb5a FIX: skip push actions whose target matches no minions
so-push-drainer dispatched every queued action, even when its target
matched no accepted minion. The orchestration's saltutil.refresh_pillar
step then failed with "No minions responded", and the drainer logged a
push failure that so-log-check reports as an ERROR.

Seen on a fresh distributed manager: the first drain after setup sent
strelka and suricata rule pushes targeting sensor roles before any
sensor had joined the grid. Nothing had failed; no node held the role.

Before dispatch, resolve each action's target with the master's
CkMinions (accepted keys and cached grains) and drop the ones that
match nothing, logging each at INFO. If none remain, clear the intents
without dispatching. Uncached minions still count as a match, and a
lookup error keeps the action, so pushes are never dropped when the
answer is uncertain.
2026-11-10 15:26:37 -05:00
Josh Patterson 9e8592e220 Merge remote-tracking branch 'origin/3/dev' into fix/auto-apply-push-result-unknown
# Conflicts:
#	salt/manager/tools/sbin/so-push-drainer
#	salt/manager/tools/sbin/so-push-drainer_test.py
2026-11-06 09:50:25 -05:00
Josh Patterson cad18f5bfc FIX: report push result as unknown when the orchestration raises
A queued push is tracked by an orchestration that waits on the minion's
state run. If salt-master restarts while it waits, the orchestration step
raises AuthenticationError and never collects the minion's return, but the
minion still runs the queued state. so-push-drainer logged these as
"push failed ... change will be applied at the next scheduled highstate",
which was wrong on both counts.

Seen on a fresh 3.4.0 standalone: hydra.enabled and telegraf.output were
saved just after SOC started a grid highstate. Both pushes queued behind
it. That highstate was the first with a license granting the vrt feature,
so salt.master wrote vrt_engine.conf and reactor_hypervisor.conf and
restarted salt-master as its last step. Both orchestrations raised, the
minion then ran both state runs with 0 failed states, and the drainer
logged two ERRORs.

When every failed step of the orchestration raised (salt's "An exception
occurred in this state:" comment, with no minion returns), log a WARNING
that the result is unknown and the state run may still have completed.
Conflicts, failed states, render errors, and results mixing a raised step
with a real failure still log ERROR.
2026-10-05 17:48:47 -04:00
25 changed files with 1961 additions and 297 deletions

No files matched your search

+2
View File
@@ -131,6 +131,8 @@ def beacon(config): # noqa: C901
'setting_id': setting_id,
'node_id': node_id,
})
log.info('postgres_pillar_beacon: audit_settings id=%d setting_id=%s node_id=%s',
row_id, setting_id, node_id)
if row_id > max_id:
max_id = row_id
+102
View File
@@ -656,6 +656,108 @@ set_version() {
fi
}
so_telegraf_output() {
local defaultsfile="${DEFAULT_SALT_DIR}/salt/telegraf/defaults.yaml"
local output
output=$(lookup_pillar output telegraf) || return 1
if [ -z "$output" ]; then
if [ ! -f "$defaultsfile" ]; then
echo "Could not find $defaultsfile" >&2
return 1
fi
output=$(grep -E "^ output: " "$defaultsfile" | awk -F: '{print $2}' | tr -d '[:space:]')
fi
echo "${output^^}"
}
so_telegraf_pg_check() {
local app="$1"
local databases role_ok
if ! docker_check_running postgres; then
echo "so-postgres is not running. Run $app on the manager." >&2
return 1
fi
if ! databases=$(docker exec so-postgres psql -U postgres -lqt 2>/dev/null); then
echo "Unable to list the so-postgres databases." >&2
return 1
fi
if ! echo "$databases" | cut -d\| -f1 | grep -qw so_telegraf; then
echo "Database so_telegraf not found. Is telegraf.output set to POSTGRES or BOTH?" >&2
return 1
fi
if ! role_ok=$(docker exec so-postgres psql -U postgres -tAc "SELECT rolcanlogin AND NOT (rolsuper OR rolcreatedb OR rolcreaterole OR rolreplication OR rolbypassrls) FROM pg_roles WHERE rolname = 'so_maint_telegraf'" 2>/dev/null); then
echo "Unable to query the so-postgres roles." >&2
return 1
fi
if [ -z "$role_ok" ]; then
echo "The so_maint_telegraf role is missing; run a highstate." >&2
return 1
elif [ "$role_ok" != t ]; then
echo "The so_maint_telegraf role is misconfigured; run a highstate." >&2
return 1
fi
}
so_telegraf_psql() {
local app="$1"
local sql="$2"
docker exec -e PGOPTIONS="-c application_name=${app} -c default_transaction_read_only=on -c client_min_messages=error" \
so-postgres psql -U so_maint_telegraf -d so_telegraf -v ON_ERROR_STOP=1 -At -F '|' -c "$sql"
}
# Prints one stream|current|then line per data stream, in bytes summed across hosts. current is
# each host's newest size in the past 4 hours; then is that host's newest size in the window
# [now - start, now - stop), or its current size when it has none there, so a host missing from the
# window adds no growth. Empty output means the window has no rows at all. Values over 15 digits
# (1 PB) are skipped so the sums here and in the callers stay far below the 64-bit limit; such a
# value can only come from a forged Telegraf writer.
so_es_index_size_growth() {
local app="$1"
local start="$2"
local stop="$3"
local relkind
relkind=$(so_telegraf_psql "$app" "SELECT c.relkind FROM pg_catalog.pg_class c JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace WHERE n.nspname = 'telegraf' AND c.relname = 'elasticsearch_index_size'") || return 1
[[ "$relkind" == "r" || "$relkind" == "p" ]] || return 0
so_telegraf_psql "$app" "$(cat <<EOF
WITH sizes AS (
SELECT d.tag_id, d.time, f.key AS stream, f.value::bigint AS bytes
FROM telegraf.elasticsearch_index_size d
CROSS JOIN LATERAL pg_catalog.jsonb_each_text(d.fields) f
WHERE d.time >= now() - interval '${start}'
AND f.key ~ '^[^|"''[:space:][:cntrl:]]+$'
AND pg_catalog.strpos(f.key, pg_catalog.chr(92)) = 0
AND f.value ~ '^[0-9]{1,15}$'
),
cur AS (
SELECT DISTINCT ON (tag_id, stream) tag_id, stream, bytes
FROM sizes
WHERE time >= now() - interval '4 hours'
ORDER BY tag_id, stream, time DESC
),
past AS (
SELECT DISTINCT ON (tag_id, stream) tag_id, stream, bytes
FROM sizes
WHERE time < now() - interval '${stop}'
ORDER BY tag_id, stream, time DESC
)
SELECT c.stream, sum(c.bytes), sum(coalesce(p.bytes, c.bytes))
FROM cur c
LEFT JOIN past p USING (tag_id, stream)
WHERE EXISTS (
SELECT 1 FROM telegraf.elasticsearch_index_size
WHERE time >= now() - interval '${start}' AND time < now() - interval '${stop}'
)
GROUP BY c.stream
ORDER BY c.stream
EOF
)"
}
status () {
printf "\n=========================================================================\n$(date) | $1\n=========================================================================\n"
}
+1
View File
@@ -244,6 +244,7 @@ if [[ $EXCLUDE_KNOWN_ERRORS == 'Y' ]]; then
EXCLUDED_ERRORS="$EXCLUDED_ERRORS|TransformTask\] \[logs-.*user so_kibana lacks the required permissions" # Known issue with integrations starting transform jobs that are explicitly not allowed to start as a system user
EXCLUDED_ERRORS="$EXCLUDED_ERRORS|manifest unknown" # so-dockerregistry logs a tag lookup miss during image copy; not tied to one docker version
EXCLUDED_ERRORS="$EXCLUDED_ERRORS|Could not index event to Elasticsearch.*\"version\" => \"9.0.8\"" # Expected during Elastic upgrade temporarily, as policies referencing older pipelines are updated
EXCLUDED_ERRORS="$EXCLUDED_ERRORS|push result unknown.*AuthenticationError" # so-push-drainer WARNING: salt-master restarted while a push was queued; the queued state run may still complete
fi
RESULT=0
-1
View File
@@ -103,7 +103,6 @@ docker:
- 80:80
- 443:443
- 8443:8443
- 7788:7788
- 7789:7789
custom_bind_mounts: []
extra_hosts: []
@@ -5,6 +5,10 @@
# https://securityonion.net/license; you may not use this file except in compliance with the
# Elastic License 2.0.
# Reports per data stream Elasticsearch growth over the last 24 hours, 7 days and 30 days. The data
# source follows telegraf.output: Postgres in POSTGRES mode, where history is limited to
# postgres.telegraf.retention_days and the script must run on the manager; InfluxDB otherwise.
INFLUX_URL="https://localhost:8086/api/v2"
. /usr/sbin/so-common
@@ -18,8 +22,6 @@ lookup_org_id() {
echo "$response" | jq -r ".orgs[] | select(.name == \"Security Onion\").id"
}
ORG_ID=$(lookup_org_id)
run_flux_query() {
local query=$1
request "query?org=$ORG_ID" -H 'Accept:application/csv' -H 'Content-type:application/vnd.flux' -d "$query" -XPOST 2>/dev/null
@@ -39,75 +41,137 @@ bytes_to_gb() {
fi
}
indexes_query='from(bucket: "telegraf/so_long_term")
|> range(start: -7d)
|> filter(fn: (r) => r._measurement == "elasticsearch_index_size")
|> distinct(column: "_field")
|> keep(columns: ["_field"])'
influxdb_indices_growth() {
ORG_ID=$(lookup_org_id)
indexes_result=$(run_flux_query "$indexes_query")
indexes=$(echo "$indexes_result" | tail -n +2 | cut -d',' -f4 | grep -v '^$' | grep -v '^_field$' | sed 's/\r$//' | sort -u)
indexes_query='from(bucket: "telegraf/so_long_term")
|> range(start: -7d)
|> filter(fn: (r) => r._measurement == "elasticsearch_index_size")
|> distinct(column: "_field")
|> keep(columns: ["_field"])'
printf "%-50s %15s %15s %15s\n" "Index Name" "Last 24hr (GB)" "Last 7d (GB)" "Last 30d (GB)"
printf "%-50s %15s %15s %15s\n" "$(printf '%.0s-' {1..50})" "$(printf '%.0s-' {1..15})" "$(printf '%.0s-' {1..15})" "$(printf '%.0s-' {1..15})"
indexes_result=$(run_flux_query "$indexes_query")
indexes=$(echo "$indexes_result" | tail -n +2 | cut -d',' -f4 | grep -v '^$' | grep -v '^_field$' | sed 's/\r$//' | sort -u)
for index in $indexes; do
[[ -z "$index" ]] && continue
current_query="from(bucket: \"telegraf/so_long_term\")
|> range(start: -4h)
|> filter(fn: (r) => r._measurement == \"elasticsearch_index_size\" and r._field == \"$index\")
|> last()
|> keep(columns: [\"_value\"])"
current_result=$(run_flux_query "$current_query")
current_size=$(read_csv_result "$current_result")
current_size=${current_size:-0}
printf "%-50s %15s %15s %15s\n" "Index Name" "Last 24hr (GB)" "Last 7d (GB)" "Last 30d (GB)"
printf "%-50s %15s %15s %15s\n" "$(printf '%.0s-' {1..50})" "$(printf '%.0s-' {1..15})" "$(printf '%.0s-' {1..15})" "$(printf '%.0s-' {1..15})"
size_24h_query="from(bucket: \"telegraf/so_long_term\")
|> range(start: -25h, stop: -23h)
|> filter(fn: (r) => r._measurement == \"elasticsearch_index_size\" and r._field == \"$index\")
|> last()
|> keep(columns: [\"_value\"])"
size_24h_result=$(run_flux_query "$size_24h_query")
size_24h_ago=$(read_csv_result "$size_24h_result")
size_24h_ago=${size_24h_ago:-$current_size}
for index in $indexes; do
[[ -z "$index" ]] && continue
current_query="from(bucket: \"telegraf/so_long_term\")
|> range(start: -4h)
|> filter(fn: (r) => r._measurement == \"elasticsearch_index_size\" and r._field == \"$index\")
|> last()
|> keep(columns: [\"_value\"])"
current_result=$(run_flux_query "$current_query")
current_size=$(read_csv_result "$current_result")
current_size=${current_size:-0}
size_7d_query="from(bucket: \"telegraf/so_long_term\")
|> range(start: -7d8h, stop: -7d)
|> filter(fn: (r) => r._measurement == \"elasticsearch_index_size\" and r._field == \"$index\")
|> last()
|> keep(columns: [\"_value\"])"
size_7d_result=$(run_flux_query "$size_7d_query")
size_7d_ago=$(read_csv_result "$size_7d_result")
size_7d_ago=${size_7d_ago:-$current_size}
size_24h_query="from(bucket: \"telegraf/so_long_term\")
|> range(start: -25h, stop: -23h)
|> filter(fn: (r) => r._measurement == \"elasticsearch_index_size\" and r._field == \"$index\")
|> last()
|> keep(columns: [\"_value\"])"
size_24h_result=$(run_flux_query "$size_24h_query")
size_24h_ago=$(read_csv_result "$size_24h_result")
size_24h_ago=${size_24h_ago:-$current_size}
size_30d_query="from(bucket: \"telegraf/so_long_term\")
|> range(start: -30d8h, stop: -30d)
|> filter(fn: (r) => r._measurement == \"elasticsearch_index_size\" and r._field == \"$index\")
|> last()
|> keep(columns: [\"_value\"])"
size_30d_result=$(run_flux_query "$size_30d_query")
size_30d_ago=$(read_csv_result "$size_30d_result")
size_30d_ago=${size_30d_ago:-$current_size}
size_7d_query="from(bucket: \"telegraf/so_long_term\")
|> range(start: -7d8h, stop: -7d)
|> filter(fn: (r) => r._measurement == \"elasticsearch_index_size\" and r._field == \"$index\")
|> last()
|> keep(columns: [\"_value\"])"
size_7d_result=$(run_flux_query "$size_7d_query")
size_7d_ago=$(read_csv_result "$size_7d_result")
size_7d_ago=${size_7d_ago:-$current_size}
# if an index was recently cleaned up by ilm it will result in a negative number for 'index growth'.
growth_24h=$(( current_size > size_24h_ago ? current_size - size_24h_ago : 0 ))
size_30d_query="from(bucket: \"telegraf/so_long_term\")
|> range(start: -30d8h, stop: -30d)
|> filter(fn: (r) => r._measurement == \"elasticsearch_index_size\" and r._field == \"$index\")
|> last()
|> keep(columns: [\"_value\"])"
size_30d_result=$(run_flux_query "$size_30d_query")
size_30d_ago=$(read_csv_result "$size_30d_result")
size_30d_ago=${size_30d_ago:-$current_size}
growth_7d=$(( current_size > size_7d_ago ? current_size - size_7d_ago : 0 ))
# if an index was recently cleaned up by ilm it will result in a negative number for 'index growth'.
growth_24h=$(( current_size > size_24h_ago ? current_size - size_24h_ago : 0 ))
growth_30d=$(( current_size > size_30d_ago ? current_size - size_30d_ago : 0 ))
growth_7d=$(( current_size > size_7d_ago ? current_size - size_7d_ago : 0 ))
growth_24h_gb=$(bytes_to_gb "$growth_24h")
growth_7d_gb=$(bytes_to_gb "$growth_7d")
growth_30d_gb=$(bytes_to_gb "$growth_30d")
growth_30d=$(( current_size > size_30d_ago ? current_size - size_30d_ago : 0 ))
# Only results for indices with atleast 1 metric above 0.00
if [[ "$growth_24h_gb" != "0.00" ]] || [[ "$growth_7d_gb" != "0.00" ]] || [[ "$growth_30d_gb" != "0.00" ]]; then
printf "%020.2f|%-50s %15s %15s %15s\n" \
"$growth_24h" \
"$index" \
"$growth_24h_gb" \
"$growth_7d_gb" \
"$growth_30d_gb"
fi
done | sort -t'|' -k1,1nr | cut -d'|' -f2-
growth_24h_gb=$(bytes_to_gb "$growth_24h")
growth_7d_gb=$(bytes_to_gb "$growth_7d")
growth_30d_gb=$(bytes_to_gb "$growth_30d")
# Only results for indices with atleast 1 metric above 0.00
if [[ "$growth_24h_gb" != "0.00" ]] || [[ "$growth_7d_gb" != "0.00" ]] || [[ "$growth_30d_gb" != "0.00" ]]; then
printf "%020.2f|%-50s %15s %15s %15s\n" \
"$growth_24h" \
"$index" \
"$growth_24h_gb" \
"$growth_7d_gb" \
"$growth_30d_gb"
fi
done | sort -t'|' -k1,1nr | cut -d'|' -f2-
}
postgres_indices_growth() {
local app="so-elasticsearch-indices-growth"
local -a starts=("25 hours" "7 days 8 hours" "30 days 8 hours")
local -a stops=("23 hours" "7 days" "30 days")
local -A current past window_has_data
local i rows stream current_size past_size growth growth_gb sort_key show
local -a columns
for i in 0 1 2; do
if ! rows=$(so_es_index_size_growth "$app" "${starts[i]}" "${stops[i]}"); then
echo "Failed to query telegraf.elasticsearch_index_size in so_telegraf." >&2
exit 1
fi
[[ -z "$rows" ]] && continue
window_has_data[$i]=1
while IFS='|' read -r stream current_size past_size; do
current[$stream]=$current_size
past[$i,$stream]=$past_size
done <<< "$rows"
done
printf "%-50s %15s %15s %15s\n" "Index Name" "Last 24hr (GB)" "Last 7d (GB)" "Last 30d (GB)"
printf "%-50s %15s %15s %15s\n" "$(printf '%.0s-' {1..50})" "$(printf '%.0s-' {1..15})" "$(printf '%.0s-' {1..15})" "$(printf '%.0s-' {1..15})"
for stream in "${!current[@]}"; do
columns=()
sort_key=0
show=false
for i in 0 1 2; do
if [[ -z "${window_has_data[$i]}" ]]; then
columns+=("n/a")
continue
fi
current_size=${current[$stream]}
past_size=${past[$i,$stream]-$current_size}
growth=$(( current_size > past_size ? current_size - past_size : 0 ))
growth_gb=$(bytes_to_gb "$growth")
[[ "$growth_gb" != "0.00" ]] && show=true
[[ $i -eq 0 ]] && sort_key=$growth
columns+=("$growth_gb")
done
if [[ "$show" == true ]]; then
printf "%020.2f|%-50s %15s %15s %15s\n" "$sort_key" "$stream" "${columns[@]}"
fi
done | sort -t'|' -k1,1nr | cut -d'|' -f2-
}
if ! TG_OUTPUT=$(so_telegraf_output); then
echo "Unable to read telegraf.output from pillar." >&2
exit 1
fi
if [[ "$TG_OUTPUT" == "POSTGRES" ]]; then
so_telegraf_pg_check so-elasticsearch-indices-growth || exit 1
postgres_indices_growth
else
influxdb_indices_growth
fi
@@ -58,6 +58,9 @@ Usage: $(basename "$0") [OPTIONS]
Estimate remaining days until Elasticsearch cluster reaches low watermark threshold.
Index growth history follows telegraf.output: in POSTGRES mode it is read from Postgres, limited to
postgres.telegraf.retention_days, and the script must run on the manager; otherwise from InfluxDB.
OPTIONS:
--json Output results in JSON format
-v, --verbose Show additional output
@@ -168,11 +171,25 @@ run_indices_growth() {
fi
}
fetch_postgres_total_bytes() {
local total
total=$(so_es_index_size_growth so-elasticsearch-retention-estimate "$1" "$2") || return 1
echo "$total" | awk -F'|' '{ sum += $3 } END { print sum + 0 }'
}
fetch_total_bytes() {
local start="$1"
local stop="$2"
local pg_start="$3"
local pg_stop="$4"
local range_line
if [ "$TG_OUTPUT" = "POSTGRES" ]; then
fetch_postgres_total_bytes "$pg_start" "$pg_stop"
return
fi
if [ -n "$stop" ]; then
range_line=" |> range(start: ${start}, stop: ${stop})"
else
@@ -208,8 +225,13 @@ fail() {
echo -e "\nDISCLAIMER: Script output is based on current data patterns, but are approximations solely intended to assist with getting a general ILM policy configured."
ORG_ID=$(lookup_org_id)
[ -n "$ORG_ID" ] || fail "Unable to resolve InfluxDB org id"
TG_OUTPUT=$(so_telegraf_output) || fail "Unable to read telegraf.output from pillar."
if [ "$TG_OUTPUT" = "POSTGRES" ]; then
pg_check_error=$(so_telegraf_pg_check so-elasticsearch-retention-estimate 2>&1) || fail "$pg_check_error"
else
ORG_ID=$(lookup_org_id)
[ -n "$ORG_ID" ] || fail "Unable to resolve InfluxDB org id"
fi
cluster_storage_size=0
indexed_storage_source="elasticsearch"
@@ -221,10 +243,11 @@ if [ -n "$cluster_storage_size_output" ]; then
fi
fi
# historical data from influxdb for growth calculation
one_day_total=$(fetch_total_bytes "-25h" "-23h")
seven_day_total=$(fetch_total_bytes "-7d8h" "-7d")
thirty_day_total=$(fetch_total_bytes "-30d8h" "-30d")
# historical index sizes for growth calculation
query_error="Failed to query telegraf.elasticsearch_index_size in so_telegraf."
one_day_total=$(fetch_total_bytes "-25h" "-23h" "25 hours" "23 hours") || fail "$query_error"
seven_day_total=$(fetch_total_bytes "-7d8h" "-7d" "7 days 8 hours" "7 days") || fail "$query_error"
thirty_day_total=$(fetch_total_bytes "-30d8h" "-30d" "30 days 8 hours" "30 days") || fail "$query_error"
# available historical windows (prefer 30d/7d when available, to avoid using recent 24h traffic spike as true daily ingest rate)
history_days=0
@@ -244,7 +267,12 @@ elif [ "$one_day_total" -gt 0 ]; then
historical_total=$one_day_total
fi
[ "$history_days" -gt 0 ] || fail "Historical InfluxDB data unavailable for growth calculation. If this a newer grid try re-running this script in a few days. Otherwise review /opt/so/log/telegraf/telegraf.log for errors with collecting required ES metrics."
if [ "$TG_OUTPUT" = "POSTGRES" ]; then
history_unavailable="Historical Postgres data unavailable for growth calculation. Postgres keeps postgres.telegraf.retention_days days of history; if this is a newer grid try re-running this script in a few days. Otherwise review /opt/so/log/telegraf/telegraf.log for errors with collecting required ES metrics."
else
history_unavailable="Historical InfluxDB data unavailable for growth calculation. If this a newer grid try re-running this script in a few days. Otherwise review /opt/so/log/telegraf/telegraf.log for errors with collecting required ES metrics."
fi
[ "$history_days" -gt 0 ] || fail "$history_unavailable"
# Daily growth rate
growth_bytes=$(( cluster_storage_size - historical_total ))
@@ -729,6 +757,7 @@ if [ "$should_trigger_recommendations" = true ]; then
[ -n "$index" ] || continue
growth_24h_gb=$(echo "$line" | awk '{print $(NF-2)}')
[[ "$growth_24h_gb" =~ ^[0-9]+(\.[0-9]+)?$ ]] || growth_24h_gb=0
creation_date_display=""
retention_days=""
-10
View File
@@ -118,10 +118,6 @@ firewall:
tcp:
- 5601
udp: []
localrules:
tcp:
- 7788
udp: []
nginx:
tcp:
- 80
@@ -200,7 +196,6 @@ firewall:
- postgres
- elasticsearch_rest
- elasticsearch_node
- localrules
sensor:
portgroups:
- beats_5044
@@ -391,7 +386,6 @@ firewall:
- elastic_agent_control
- elastic_agent_data
- elastic_agent_update
- localrules
- sensoroni
fleet:
portgroups:
@@ -611,7 +605,6 @@ firewall:
- elastic_agent_control
- elastic_agent_data
- elastic_agent_update
- localrules
- sensoroni
fleet:
portgroups:
@@ -829,7 +822,6 @@ firewall:
- elastic_agent_control
- elastic_agent_data
- elastic_agent_update
- localrules
- sensoroni
fleet:
portgroups:
@@ -1058,7 +1050,6 @@ firewall:
- elastic_agent_update
- endgame
- strelka_frontend
- localrules
fleet:
portgroups:
- elasticsearch_rest
@@ -1518,7 +1509,6 @@ firewall:
- elasticsearch_rest
- elasticsearch_node
- elastic_agent_control
- localrules
sensor:
portgroups:
- beats_5044
-3
View File
@@ -138,9 +138,6 @@ firewall:
kibana:
tcp: *tcpsettings
udp: *udpsettings
localrules:
tcp: *tcpsettings
udp: *udpsettings
nginx:
tcp: *tcpsettings
udp: *udpsettings
+245 -6
View File
@@ -16,9 +16,12 @@ For each intent file under /opt/so/state/push_pending/*.json whose last_touch
is older than debounce_seconds, this script:
* concatenates the actions lists from every ready intent
* dedupes by (state or __highstate__, tgt, tgt_type)
* drops actions whose target matches no accepted minion (e.g. a role with no nodes yet)
* dispatches a single `salt-run state.orchestrate orch.push_batch --async`
with the deduped actions list passed as pillar kwargs
* deletes the contributed intent files on successful dispatch
* records the orchestration jid under /opt/so/state/push_dispatched and, on
later passes, looks up its result and logs success or per-minion failures
Reactor sls files (push_files, push_pillar) write intents
but never dispatch directly
@@ -30,18 +33,36 @@ import json
import logging
import logging.handlers
import os
import re
import subprocess
import sys
import time
import salt.client
import salt.config
import salt.utils.minions
PENDING_DIR = '/opt/so/state/push_pending'
LOCK_FILE = os.path.join(PENDING_DIR, '.lock')
LOG_FILE = '/opt/so/log/salt/so-push-drainer.log'
MASTER_CONFIG = '/etc/salt/master'
DISPATCHED_DIR = '/opt/so/state/push_dispatched'
HIGHSTATE_SENTINEL = '__highstate__'
RESULT_CHECK_DELAY = 30
RESULT_RECHECK_MAX = 300
RESULT_MAX_AGE = 7200
RESULT_CHECKS_PER_PASS = 5
TEXT_LIMIT = 500
# Lead-in salt puts on the comment of an orchestration step that raised.
STEP_RAISED = 'An exception occurred in this state:'
# salt-run --async reports the jid only in a log line (stderr by default).
JID_RE = re.compile(r'salt/run/(\d{20})')
def _make_logger():
logger = logging.getLogger('so-push-drainer')
@@ -96,6 +117,29 @@ def _dedupe_actions(actions):
return deduped
def _targeted(actions, log):
"""Drops actions whose target matches no accepted minion (e.g. a role with no nodes yet), which
the orchestration would otherwise report as failed with 'No minions responded'."""
try:
ckminions = salt.utils.minions.CkMinions(salt.config.master_config(MASTER_CONFIG))
except Exception:
log.exception('cannot load master config; dispatching without checking targets')
return actions
targeted = []
for action in actions:
try:
matched = ckminions.check_minions(action['tgt'], action.get('tgt_type', 'compound')).get('minions')
except Exception:
log.exception('cannot resolve tgt=%s; dispatching it anyway', action.get('tgt'))
matched = True
if matched:
targeted.append(action)
else:
log.info('skipping %s: no minions match tgt=%s',
'highstate' if action.get('highstate') else action.get('state'), action.get('tgt'))
return targeted
def _dispatch(actions, log):
pillar_arg = json.dumps({'actions': actions})
cmd = [
@@ -113,14 +157,196 @@ def _dispatch(actions, log):
except subprocess.CalledProcessError as exc:
log.error('dispatch failed (rc=%s): stdout=%s stderr=%s',
exc.returncode, exc.stdout, exc.stderr)
return False
return None
except subprocess.TimeoutExpired:
log.error('dispatch timed out after 60s')
return False
return None
except Exception:
log.exception('dispatch raised')
return None
output = '{}\n{}'.format(result.stderr or '', result.stdout or '')
match = JID_RE.search(output)
if not match:
log.warning('dispatch accepted but no jid found, result will not be tracked: output=%s',
_trim(output))
return ''
log.info('dispatch accepted: jid=%s', match.group(1))
return match.group(1)
def _trim(value):
text = value if isinstance(value, str) else json.dumps(value, default=str)
lines = [line.strip() for line in text.splitlines() if line.strip()]
if 'Traceback (most recent call last):' in text:
# Keep the lead-in and the raised exception; the frames are noise in a log line.
lines = [text.split('Traceback (most recent call last):', 1)[0].strip(), lines[-1]]
text = ' '.join(line for line in lines if line)
return text if len(text) <= TEXT_LIMIT else text[:TEXT_LIMIT] + '...'
def _unlink(path, log):
try:
os.unlink(path)
except FileNotFoundError:
pass
except OSError:
log.exception('failed to remove %s', path)
def _write_record(path, record, log):
try:
os.makedirs(DISPATCHED_DIR, exist_ok=True)
tmp_path = path + '.tmp'
with open(tmp_path, 'w') as f:
json.dump(record, f)
os.rename(tmp_path, path)
except Exception:
log.exception('failed to record dispatch %s', record.get('jid'))
def _record_dispatch(jid, actions, paths, log):
record = {'jid': jid, 'dispatched_at': time.time(), 'actions': actions, 'paths': paths}
_write_record(os.path.join(DISPATCHED_DIR, '{}.json'.format(jid)), record, log)
def _lookup_jid(jid, log):
"""Returns the job cache entry for jid, {} while it is still running, or None on error."""
cmd = ['salt-run', 'jobs.lookup_jid', jid, '--out=json']
try:
result = subprocess.run(cmd, check=True, capture_output=True, text=True, timeout=60)
return json.loads(result.stdout or '{}')
except (subprocess.CalledProcessError, subprocess.TimeoutExpired, ValueError) as exc:
log.warning('lookup of jid %s failed: %s', jid, exc)
return None
def _minion_failure(minion_ret):
if isinstance(minion_ret, dict):
return '; '.join(
'{}: {}'.format(state.get('__id__', state_key), _trim(state.get('comment', '')))
for state_key, state in minion_ret.items()
if isinstance(state, dict) and state.get('result') is False
)
# A state run rejected before it starts (e.g. another state run is in
# progress) returns a list of error strings instead of state results.
if isinstance(minion_ret, (list, str)):
return _trim(minion_ret)
return ''
def _step_failures(step):
if not isinstance(step, dict) or step.get('result') is not False:
return []
failures = ['{}: {}'.format(step.get('__id__', step.get('name')), _trim(step.get('comment', '')))]
changes = step.get('changes')
minion_rets = changes.get('ret') if isinstance(changes, dict) else None
if isinstance(minion_rets, dict):
for minion, minion_ret in minion_rets.items():
text = _minion_failure(minion_ret)
if text:
failures.append('{}: {}'.format(minion, text))
return failures
def _orch_failures(ret):
if not isinstance(ret, dict):
return [_trim(ret)]
failures = []
for job in ret.values():
if not isinstance(job, dict):
continue
job_ret = job.get('return')
data = job_ret.get('data') if isinstance(job_ret, dict) else {}
if not isinstance(data, dict):
if data:
failures.append(_trim(data))
data = {}
for steps in data.values():
if not isinstance(steps, dict):
failures.append(_trim(steps))
continue
for step in steps.values():
failures.extend(_step_failures(step))
if job.get('success') is False and not failures:
failures.append('orchestration reported failure: {}'.format(_trim(job.get('return'))))
return failures
def _failed_steps(ret):
steps = []
for job in ret.values() if isinstance(ret, dict) else []:
job_ret = job.get('return') if isinstance(job, dict) else None
data = job_ret.get('data') if isinstance(job_ret, dict) else None
for group in data.values() if isinstance(data, dict) else []:
if isinstance(group, dict):
steps.extend(step for step in group.values() if isinstance(step, dict) and step.get('result') is False)
return steps
def _result_unknown(ret):
# A step that raised (e.g. salt-master restarted while it waited on a queued state run)
# never collected the minion's return, so the state run may still have completed.
steps = _failed_steps(ret)
return bool(steps) and all(str(step.get('comment', '')).startswith(STEP_RAISED) for step in steps)
def _recheck_delay(age):
return min(RESULT_RECHECK_MAX, max(RESULT_CHECK_DELAY, age / 4))
def _check_dispatched(log, now):
due = []
for path in glob.glob(os.path.join(DISPATCHED_DIR, '*.json')):
record = _read_intent(path, log)
if not isinstance(record, dict) or not record.get('jid'):
_unlink(path, log)
continue
try:
age = now - record.get('dispatched_at', 0)
last_check = record.get('checked_at', record.get('dispatched_at', 0))
is_due = now - last_check >= _recheck_delay(age)
except TypeError:
log.warning('malformed record for jid=%s; no longer tracking', record.get('jid'))
_unlink(path, log)
continue
if is_due:
due.append((last_check, path, record, age))
# Least recently checked first, so pushes that are still running can't starve finished ones.
for _, path, record, age in sorted(due, key=lambda item: item[:2])[:RESULT_CHECKS_PER_PASS]:
jid = record['jid']
try:
if _report_result(record, age, log):
_unlink(path, log)
else:
record['checked_at'] = now
_write_record(path, record, log)
except Exception:
# Drop the record so one unreadable result can't fail every pass ahead of the drain.
log.exception('cannot evaluate result for jid=%s; no longer tracking', jid)
_unlink(path, log)
def _report_result(record, age, log):
"""Logs the outcome of a dispatched push. Returns True once the record is finished with."""
jid = record['jid']
paths = record.get('paths', [])
ret = _lookup_jid(jid, log)
if not ret:
if age > RESULT_MAX_AGE:
log.warning('no result for jid=%s after %ds, no longer tracking; paths=%s', jid, age, paths)
return True
return False
log.info('dispatch accepted: %s', (result.stdout or '').strip())
failures = _orch_failures(ret)
if failures and _result_unknown(ret):
# so-log-check excludes this message by its text.
log.warning('push result unknown jid=%s paths=%s; the orchestration lost track of the state run, which may '
'still have completed (if not, the change will be applied at the next scheduled highstate): %s',
jid, paths, ' | '.join(failures))
elif failures:
log.error('push failed jid=%s paths=%s; change will be applied at the next scheduled highstate: %s',
jid, paths, ' | '.join(failures))
else:
log.info('push succeeded jid=%s paths=%s', jid, paths)
return True
@@ -143,6 +369,9 @@ def main():
debounce_seconds = int(push.get('debounce_seconds', 30))
# Outside the lock: lookups are slow and the reactors take the same lock.
_check_dispatched(log, time.time())
os.makedirs(PENDING_DIR, exist_ok=True)
lock_fd = os.open(LOCK_FILE, os.O_CREAT | os.O_RDWR, 0o644)
try:
@@ -208,10 +437,20 @@ def main():
len(ready), len(deduped), len(combined_actions),
debounce_duration, all_paths[:20],
)
for action in deduped:
log.info('action: %s tgt=%s', 'highstate' if action.get('highstate') else action.get('state'),
action.get('tgt'))
if not _dispatch(deduped, log):
log.warning('dispatch failed; leaving intent files in place for retry')
return 1
targeted = _targeted(deduped, log)
if targeted:
jid = _dispatch(targeted, log)
if jid is None:
log.warning('dispatch failed; leaving intent files in place for retry')
return 1
if jid:
_record_dispatch(jid, targeted, all_paths[:20], log)
else:
log.info('no action matches a minion; nothing to dispatch')
for path, _ in ready:
try:
@@ -0,0 +1,585 @@
# Copyright Security Onion Solutions LLC and/or licensed to Security Onion Solutions LLC under one
# or more contributor license agreements. Licensed under the Elastic License 2.0 as shown at
# https://securityonion.net/license; you may not use this file except in compliance with the
# Elastic License 2.0.
import importlib.util
import json
import logging
import os
import shutil
import subprocess
import sys
import tempfile
import time
import unittest
from importlib.machinery import SourceFileLoader
from unittest.mock import MagicMock, patch
HERE = os.path.dirname(os.path.abspath(__file__))
SCRIPT = os.path.join(HERE, 'so-push-drainer')
_loader = SourceFileLoader('so_push_drainer', SCRIPT)
_spec = importlib.util.spec_from_loader('so_push_drainer', _loader)
drainer = importlib.util.module_from_spec(_spec)
# salt is not installed where these tests run; the drainer only needs salt.client.Caller and CkMinions.
# Mocked only while the drainer loads: run from the repo root, 'salt' is this repo's salt/ directory.
_salt = MagicMock()
with patch.dict(sys.modules, {'salt': _salt, 'salt.client': _salt.client, 'salt.config': _salt.config,
'salt.utils': _salt.utils, 'salt.utils.minions': _salt.utils.minions}):
_loader.exec_module(drainer)
MASTER = 'manager.localdomain_master'
JID = '20260930171554259426'
ASYNC_STDERR = ('[WARNING ] Running in asynchronous mode. Results of this execution may be collected '
'by attaching to the master event bus or by examining the master job cache, if '
'configured. This execution is running under tag salt/run/{}\n'.format(JID))
CONFLICT = ('The function "state.sls" is running as PID 372218 and was started at '
'2026, Sep 30 17:15:40.466233 with jid 20260930171540466233')
def _orch_ret(steps, success=True):
return {MASTER: {
'fun': 'runner.state.orchestrate',
'jid': JID,
'return': {'data': {MASTER: steps}, 'outputter': 'highstate', 'retcode': 0 if success else 1},
'success': success,
}}
REFRESH_STEP = {
'salt_|-refresh_pillar_1_|-saltutil.refresh_pillar_|-function': {
'__id__': 'refresh_pillar_1', 'result': True,
'changes': {'ret': {'manager_standalone': True}},
'comment': 'Function ran successfully.',
},
}
CONFLICT_RET = _orch_ret(dict(REFRESH_STEP, **{
'salt_|-apply_soc_1_|-apply_soc_1_|-state': {
'__id__': 'apply_soc_1', 'result': False,
'changes': {'out': 'highstate', 'ret': {'manager_standalone': [CONFLICT]}},
'comment': 'Run failed on minions: manager_standalone',
},
}), success=False)
STATE_FAIL_RET = _orch_ret({
'salt_|-apply_hydra_1_|-apply_hydra_1_|-state': {
'__id__': 'apply_hydra_1', 'result': False,
'changes': {'out': 'highstate', 'ret': {'manager_standalone': {
'test_|-no_license_|-no_license_|-fail_without_changes': {
'__id__': 'hydra.enabled_no_license_detected', 'result': False,
'comment': 'This is a feature supported only for customers with a valid license.',
},
'file_|-hydra_conf_|-/opt/so/conf/hydra_|-managed': {'result': True, 'comment': 'ok'},
}}},
'comment': 'Run failed on minions: manager_standalone',
},
}, success=False)
RAISED = ('An exception occurred in this state: Traceback (most recent call last):\n'
' File "salt/client/__init__.py", line 1934, in pub\n'
' raise AuthenticationError(err_msg)\n'
'salt.exceptions.AuthenticationError: Authentication error occurred.\n')
RAISED_RET = _orch_ret(dict(REFRESH_STEP, **{
'salt_|-apply_hydra_1_|-apply_hydra_1_|-state': {
'__id__': 'apply_hydra_1', 'result': False, 'changes': {}, 'comment': RAISED,
},
}), success=False)
SUCCESS_RET = _orch_ret(dict(REFRESH_STEP, **{
'salt_|-apply_telegraf_1_|-apply_telegraf_1_|-state': {
'__id__': 'apply_telegraf_1', 'result': True,
'changes': {'out': 'highstate', 'ret': {'manager_standalone': {
'file_|-tgrafconf_|-/opt/so/conf/telegraf/etc/telegraf.conf_|-managed': {'result': True},
}}},
'comment': 'States ran successfully.',
},
}))
class DrainerTestCase(unittest.TestCase):
def setUp(self):
self.tmpdir = tempfile.mkdtemp()
self.pending = os.path.join(self.tmpdir, 'push_pending')
self.dispatched = os.path.join(self.tmpdir, 'push_dispatched')
os.makedirs(self.pending)
for name, value in (
('PENDING_DIR', self.pending),
('LOCK_FILE', os.path.join(self.pending, '.lock')),
('DISPATCHED_DIR', self.dispatched),
('LOG_FILE', os.path.join(self.tmpdir, 'log', 'so-push-drainer.log')),
):
patcher = patch.object(drainer, name, value)
patcher.start()
self.addCleanup(patcher.stop)
self.log = MagicMock()
def tearDown(self):
shutil.rmtree(self.tmpdir, ignore_errors=True)
def write_json(self, directory, name, data):
os.makedirs(directory, exist_ok=True)
path = os.path.join(directory, name)
with open(path, 'w') as f:
if isinstance(data, str):
f.write(data)
else:
json.dump(data, f)
return path
def logged(self, level):
return ' '.join(c.args[0] % c.args[1:] for c in getattr(self.log, level).call_args_list)
class TestHelpers(DrainerTestCase):
def test_make_logger_adds_handler_once(self):
logger = logging.getLogger('so-push-drainer')
def close_handlers():
for handler in logger.handlers:
handler.close()
logger.handlers.clear()
self.addCleanup(close_handlers)
logger.handlers.clear()
self.assertIs(drainer._make_logger(), logger)
drainer._make_logger()
self.assertEqual(len(logger.handlers), 1)
self.assertTrue(os.path.isdir(os.path.dirname(drainer.LOG_FILE)))
def test_load_push_cfg(self):
with patch.object(drainer.salt.client, 'Caller') as caller:
caller.return_value.cmd.return_value = {'enabled': False}
self.assertEqual(drainer._load_push_cfg(), {'enabled': False})
caller.return_value.cmd.return_value = 'garbage'
self.assertEqual(drainer._load_push_cfg(), {})
def test_read_intent(self):
good = self.write_json(self.pending, 'good.json', {'a': 1})
bad = self.write_json(self.pending, 'bad.json', '{nope')
self.assertEqual(drainer._read_intent(good, self.log), {'a': 1})
self.assertIsNone(drainer._read_intent(bad, self.log))
with patch('builtins.open', side_effect=RuntimeError('boom')):
self.assertIsNone(drainer._read_intent(good, self.log))
self.log.exception.assert_called_once()
def test_dedupe_actions(self):
actions = [
'not a dict',
{'state': 'soc'},
{'state': 'soc', 'tgt': '*'},
{'state': 'soc', 'tgt': '*', 'tgt_type': 'compound'},
{'highstate': True, 'tgt': '*'},
{'state': 'soc', 'tgt': 'node1', 'tgt_type': 'glob'},
]
self.assertEqual(drainer._dedupe_actions(actions), [actions[2], actions[4], actions[5]])
def test_trim(self):
self.assertEqual(drainer._trim(' text \n'), 'text')
self.assertEqual(drainer._trim(['a']), '["a"]')
self.assertEqual(drainer._trim(None), 'null')
self.assertEqual(drainer._trim('x' * 600), 'x' * drainer.TEXT_LIMIT + '...')
def test_trim_traceback(self):
comment = ('An exception occurred in this state: Traceback (most recent call last):\n'
' File "salt/client/__init__.py", line 1934, in pub\n'
' raise AuthenticationError(err_msg)\n'
'salt.exceptions.AuthenticationError: Authentication error occurred.\n')
self.assertEqual(drainer._trim(comment), 'An exception occurred in this state: '
'salt.exceptions.AuthenticationError: Authentication error occurred.')
self.assertEqual(drainer._trim('line one\n line two\n'), 'line one line two')
def test_unlink(self):
drainer._unlink(os.path.join(self.tmpdir, 'missing'), self.log)
self.log.exception.assert_not_called()
drainer._unlink(self.tmpdir, self.log)
self.log.exception.assert_called_once()
class TestDispatch(DrainerTestCase):
def run_dispatch(self, **kwargs):
with patch.object(drainer.subprocess, 'run', **kwargs) as run:
jid = drainer._dispatch([{'state': 'soc', 'tgt': '*'}], self.log)
return jid, run
def test_jid_parsed_from_stderr(self):
jid, run = self.run_dispatch(return_value=MagicMock(stdout='', stderr=ASYNC_STDERR))
self.assertEqual(jid, JID)
cmd = run.call_args[0][0]
self.assertEqual(cmd[:3], ['salt-run', 'state.orchestrate', 'orch.push_batch'])
self.assertIn('--async', cmd)
def test_jid_parsed_from_stdout(self):
jid, _ = self.run_dispatch(return_value=MagicMock(stdout=ASYNC_STDERR, stderr=None))
self.assertEqual(jid, JID)
def test_no_jid(self):
jid, _ = self.run_dispatch(return_value=MagicMock(stdout='unexpected output', stderr=None))
self.assertEqual(jid, '')
self.assertIn('output=unexpected output', self.logged('warning'))
def test_failures_return_none(self):
for exc in (subprocess.CalledProcessError(1, 'salt-run', 'out', 'err'),
subprocess.TimeoutExpired('salt-run', 60),
RuntimeError('boom')):
jid, _ = self.run_dispatch(side_effect=exc)
self.assertIsNone(jid)
def test_record_dispatch(self):
drainer._record_dispatch(JID, [{'state': 'soc'}], ['audit:soc.config.licenseKey'], self.log)
with open(os.path.join(self.dispatched, JID + '.json')) as f:
record = json.load(f)
self.assertEqual(record['jid'], JID)
self.assertEqual(record['paths'], ['audit:soc.config.licenseKey'])
self.assertIn('dispatched_at', record)
def test_record_dispatch_errors(self):
with patch.object(drainer.os, 'makedirs', side_effect=OSError('ro')):
drainer._record_dispatch(JID, [], [], self.log)
drainer._record_dispatch(JID, [object()], [], self.log)
self.assertEqual(self.log.exception.call_count, 2)
self.assertFalse(os.path.exists(os.path.join(self.dispatched, JID + '.json')))
class TestResults(DrainerTestCase):
def test_lookup_jid(self):
with patch.object(drainer.subprocess, 'run') as run:
run.return_value = MagicMock(stdout=json.dumps(SUCCESS_RET))
self.assertEqual(drainer._lookup_jid(JID, self.log), SUCCESS_RET)
self.assertEqual(run.call_args[0][0], ['salt-run', 'jobs.lookup_jid', JID, '--out=json'])
run.return_value = MagicMock(stdout='')
self.assertEqual(drainer._lookup_jid(JID, self.log), {})
run.return_value = MagicMock(stdout='not json')
self.assertIsNone(drainer._lookup_jid(JID, self.log))
run.side_effect = subprocess.TimeoutExpired('salt-run', 60)
self.assertIsNone(drainer._lookup_jid(JID, self.log))
def test_minion_failure_shapes(self):
self.assertEqual(drainer._minion_failure([CONFLICT]), json.dumps([CONFLICT]))
self.assertEqual(drainer._minion_failure('Rendering SLS failed'), 'Rendering SLS failed')
self.assertEqual(drainer._minion_failure(True), '')
self.assertEqual(drainer._minion_failure({'a': {'result': True}}), '')
def test_orch_failures_conflict(self):
failures = drainer._orch_failures(CONFLICT_RET)
self.assertEqual(failures[0], 'apply_soc_1: Run failed on minions: manager_standalone')
self.assertIn('manager_standalone', failures[1])
self.assertIn('is running as PID 372218', failures[1])
self.assertEqual(len(failures), 2)
def test_orch_failures_failed_state(self):
failures = drainer._orch_failures(STATE_FAIL_RET)
self.assertEqual(len(failures), 2)
self.assertIn('hydra.enabled_no_license_detected: This is a feature', failures[1])
self.assertNotIn('hydra_conf', failures[1])
def test_orch_failures_success(self):
self.assertEqual(drainer._orch_failures(SUCCESS_RET), [])
def test_orch_failures_render_error(self):
ret = {MASTER: {'return': {'data': {MASTER: ['Rendering SLS failed']}}, 'success': False}}
self.assertEqual(drainer._orch_failures(ret), ['["Rendering SLS failed"]'])
def test_orch_failures_not_a_dict(self):
self.assertEqual(drainer._orch_failures(['No minions matched']), ['["No minions matched"]'])
self.assertEqual(drainer._orch_failures('Runner error'), ['Runner error'])
def test_orch_failures_data_not_a_dict(self):
ret = {MASTER: {'return': {'data': ["Rendering SLS 'orch.push_batch' failed"]}, 'success': False}}
self.assertEqual(drainer._orch_failures(ret), ['["Rendering SLS \'orch.push_batch\' failed"]'])
def test_orch_failures_odd_changes(self):
for changes in ('Run failed', {'ret': ['manager_standalone']}):
ret = _orch_ret({'salt_|-apply_soc_1_|-apply_soc_1_|-state': {
'__id__': 'apply_soc_1', 'result': False, 'changes': changes, 'comment': 'Run failed on minions',
}}, success=False)
self.assertEqual(drainer._orch_failures(ret), ['apply_soc_1: Run failed on minions'])
def test_orch_failures_unparsed(self):
self.assertEqual(drainer._orch_failures({MASTER: 'odd'}), [])
ret = {MASTER: {'return': 'Exception occurred', 'success': False}}
self.assertEqual(drainer._orch_failures(ret), ['orchestration reported failure: Exception occurred'])
def test_result_unknown(self):
self.assertTrue(drainer._result_unknown(RAISED_RET))
for ret in (CONFLICT_RET, STATE_FAIL_RET, SUCCESS_RET, ['No minions matched'], {MASTER: 'odd'},
{MASTER: {'return': {'data': {MASTER: ['Rendering SLS failed']}}, 'success': False}}):
self.assertFalse(drainer._result_unknown(ret), ret)
mixed = _orch_ret(dict(RAISED_RET[MASTER]['return']['data'][MASTER],
**CONFLICT_RET[MASTER]['return']['data'][MASTER]), success=False)
self.assertFalse(drainer._result_unknown(mixed))
def record(self, jid, age, now):
return self.write_json(self.dispatched, jid + '.json', {
'jid': jid, 'dispatched_at': now - age, 'actions': [], 'paths': ['audit:' + jid],
})
def test_check_dispatched(self):
now = time.time()
results = {
'1_failed': CONFLICT_RET,
'2_ok': SUCCESS_RET,
'3_pending': {},
'4_expired': None,
'6_unknown': RAISED_RET,
}
young = self.record('0_young', 5, now)
paths = {jid: self.record(jid, 60, now) for jid in results}
paths['4_expired'] = self.record('4_expired', drainer.RESULT_MAX_AGE + 1, now)
bad = self.write_json(self.dispatched, '5_bad.json', '{nope')
with patch.object(drainer, '_lookup_jid', side_effect=lambda jid, log: results[jid]):
drainer._check_dispatched(self.log, now)
self.assertTrue(os.path.exists(young))
with open(paths['3_pending']) as f:
self.assertEqual(json.load(f)['checked_at'], now)
for jid in ('1_failed', '2_ok', '4_expired', '6_unknown'):
self.assertFalse(os.path.exists(paths[jid]), jid)
self.assertFalse(os.path.exists(bad))
self.assertIn('push failed jid=1_failed', self.logged('error'))
self.assertIn('is running as PID 372218', self.logged('error'))
self.assertIn('push succeeded jid=2_ok', self.logged('info'))
self.assertIn('no result for jid=4_expired', self.logged('warning'))
self.assertIn('push result unknown jid=6_unknown', self.logged('warning'))
self.assertIn('AuthenticationError: Authentication error occurred.', self.logged('warning'))
self.assertNotIn('6_unknown', self.logged('error'))
def test_check_dispatched_survives_bad_result(self):
now = time.time()
bad = self.record('1_bad', 60, now)
good = self.record('2_ok', 60, now)
def orch_failures(ret):
if ret == 'boom':
raise ValueError('unexpected shape')
return []
with patch.object(drainer, '_lookup_jid', side_effect=lambda jid, log: 'boom' if jid == '1_bad' else SUCCESS_RET), \
patch.object(drainer, '_orch_failures', side_effect=orch_failures):
drainer._check_dispatched(self.log, now)
self.assertFalse(os.path.exists(bad))
self.assertFalse(os.path.exists(good))
self.log.exception.assert_called_once()
self.assertIn('jid=1_bad', self.log.exception.call_args[0][0] % self.log.exception.call_args[0][1:])
self.assertIn('push succeeded jid=2_ok', self.logged('info'))
def test_check_dispatched_drops_malformed_record(self):
now = time.time()
bad_dispatched = self.write_json(self.dispatched, '1_bad.json', {'jid': '1_bad', 'dispatched_at': 'soon'})
bad_checked = self.write_json(self.dispatched, '2_bad.json', {
'jid': '2_bad', 'dispatched_at': now - 60, 'checked_at': None,
})
good = self.record('3_ok', 60, now)
with patch.object(drainer, '_lookup_jid', return_value=SUCCESS_RET) as lookup:
drainer._check_dispatched(self.log, now)
for path in (bad_dispatched, bad_checked, good):
self.assertFalse(os.path.exists(path), path)
self.assertEqual([c.args[0] for c in lookup.call_args_list], ['3_ok'])
self.assertIn('malformed record for jid=1_bad', self.logged('warning'))
self.assertIn('malformed record for jid=2_bad', self.logged('warning'))
self.assertIn('push succeeded jid=3_ok', self.logged('info'))
def test_recheck_delay(self):
self.assertEqual(drainer._recheck_delay(10), drainer.RESULT_CHECK_DELAY)
self.assertEqual(drainer._recheck_delay(400), 100)
self.assertEqual(drainer._recheck_delay(drainer.RESULT_MAX_AGE), drainer.RESULT_RECHECK_MAX)
def test_check_dispatched_limit_rotates(self):
now = time.time()
limit = drainer.RESULT_CHECKS_PER_PASS
jids = ['{:02d}'.format(i) for i in range(limit + 2)]
for jid in jids:
self.record(jid, 60, now)
with patch.object(drainer, '_lookup_jid', return_value={}) as lookup:
drainer._check_dispatched(self.log, now)
self.assertEqual([c.args[0] for c in lookup.call_args_list], jids[:limit])
lookup.reset_mock()
drainer._check_dispatched(self.log, now + 40)
self.assertEqual([c.args[0] for c in lookup.call_args_list], jids[limit:] + jids[:limit - 2])
def test_check_dispatched_not_blocked_by_running(self):
now = time.time()
for i in range(drainer.RESULT_CHECKS_PER_PASS):
self.record('1_running{}'.format(i), 600, now)
done = self.record('2_done', 60, now)
def lookup(jid, log):
return SUCCESS_RET if jid == '2_done' else {}
with patch.object(drainer, '_lookup_jid', side_effect=lookup) as lookup_jid:
drainer._check_dispatched(self.log, now)
self.assertTrue(os.path.exists(done))
lookup_jid.reset_mock()
drainer._check_dispatched(self.log, now + 15)
self.assertEqual([c.args[0] for c in lookup_jid.call_args_list], ['2_done'])
self.assertFalse(os.path.exists(done))
self.assertIn('push succeeded jid=2_done', self.logged('info'))
class TestTargeted(DrainerTestCase):
def setUp(self):
super().setUp()
self.ckminions = MagicMock()
patcher = patch.object(drainer.salt.utils.minions, 'CkMinions', return_value=self.ckminions)
self.CkMinions = patcher.start()
self.addCleanup(patcher.stop)
def test_drops_unmatched(self):
matches = {'G@role:so-sensor': ['sensor_sensor'], 'manager_manager': ['manager_manager']}
self.ckminions.check_minions.side_effect = lambda tgt, tgt_type: {'minions': matches.get(tgt, [])}
actions = [
{'state': 'strelka', 'tgt': 'G@role:so-sensor'},
{'state': 'suricata', 'tgt': 'G@role:so-heavynode'},
{'highstate': True, 'tgt': 'G@role:so-idh'},
{'state': 'soc', 'tgt': 'manager_manager', 'tgt_type': 'list'},
]
self.assertEqual(drainer._targeted(actions, self.log), [actions[0], actions[3]])
self.CkMinions.assert_called_once()
self.assertEqual([c.args for c in self.ckminions.check_minions.call_args_list], [
('G@role:so-sensor', 'compound'), ('G@role:so-heavynode', 'compound'),
('G@role:so-idh', 'compound'), ('manager_manager', 'list'),
])
info = self.logged('info')
self.assertIn('skipping suricata: no minions match tgt=G@role:so-heavynode', info)
self.assertIn('skipping highstate: no minions match tgt=G@role:so-idh', info)
self.assertNotIn('strelka', info)
def test_master_config_error_keeps_actions(self):
actions = [{'state': 'soc', 'tgt': '*'}]
with patch.object(drainer.salt.config, 'master_config', side_effect=OSError('no config')):
self.assertEqual(drainer._targeted(actions, self.log), actions)
self.log.exception.assert_called_once()
self.CkMinions.assert_not_called()
def test_check_error_keeps_action(self):
actions = [{'state': 'soc', 'tgt': 'G@role:'}]
self.ckminions.check_minions.side_effect = RuntimeError('bad target')
self.assertEqual(drainer._targeted(actions, self.log), actions)
self.log.exception.assert_called_once()
class TestMain(DrainerTestCase):
def setUp(self):
super().setUp()
self.cfg = {'enabled': True, 'debounce_seconds': 30}
for name, kwargs in (
('_make_logger', {'return_value': self.log}),
('_load_push_cfg', {'side_effect': lambda: self.cfg}),
('_check_dispatched', {}),
('_targeted', {'side_effect': lambda actions, log: actions}),
):
patcher = patch.object(drainer, name, **kwargs)
setattr(self, name, patcher.start())
self.addCleanup(patcher.stop)
def intent(self, name, age=60, actions=None, paths=None):
now = time.time()
return self.write_json(self.pending, name, {
'first_touch': now - age - 5, 'last_touch': now - age,
'actions': [{'state': 'soc', 'tgt': '*'}] if actions is None else actions,
'paths': paths or ['audit:soc.config.licenseKey'],
})
def test_no_pending_dir(self):
shutil.rmtree(self.pending)
self.assertEqual(drainer.main(), 0)
self._load_push_cfg.assert_not_called()
def test_cfg_error(self):
self._load_push_cfg.side_effect = RuntimeError('no salt')
self.assertEqual(drainer.main(), 1)
def test_disabled(self):
self.cfg['enabled'] = False
self.assertEqual(drainer.main(), 0)
self._check_dispatched.assert_not_called()
def test_no_intents_still_checks_results(self):
self.assertEqual(drainer.main(), 0)
self._check_dispatched.assert_called_once()
def test_debounce_and_broken(self):
young = self.intent('young.json', age=1)
broken = self.write_json(self.pending, 'broken.json', '{nope')
with patch.object(drainer, '_dispatch') as dispatch:
self.assertEqual(drainer.main(), 0)
dispatch.assert_not_called()
self.assertTrue(os.path.exists(young))
self.assertFalse(os.path.exists(broken))
def test_broken_unlink_error_ignored(self):
self.write_json(self.pending, 'broken.json', '{nope')
with patch.object(drainer.os, 'unlink', side_effect=OSError('busy')):
self.assertEqual(drainer.main(), 0)
def test_no_usable_actions(self):
path = self.intent('empty.json', actions=[{'state': 'soc'}])
self.assertEqual(drainer.main(), 0)
self.assertFalse(os.path.exists(path))
self.intent('empty.json', actions=[{'state': 'soc'}])
with patch.object(drainer.os, 'unlink', side_effect=OSError('busy')):
self.assertEqual(drainer.main(), 0)
def test_dispatch_failure_keeps_intents(self):
path = self.intent('pillar_soc.json')
with patch.object(drainer, '_dispatch', return_value=None):
self.assertEqual(drainer.main(), 1)
self.assertTrue(os.path.exists(path))
def test_dispatch_records_jid(self):
soc = self.intent('pillar_soc.json')
hs = self.intent('pillar_global.json', actions=[{'highstate': True, 'tgt': '*'}], paths=['audit:global.x'])
with patch.object(drainer, '_dispatch', return_value=JID) as dispatch, \
patch.object(drainer, '_record_dispatch') as record:
self.assertEqual(drainer.main(), 0)
self.assertEqual(len(dispatch.call_args[0][0]), 2)
record.assert_called_once()
self.assertEqual(record.call_args[0][0], JID)
self.assertEqual(sorted(record.call_args[0][2]), ['audit:global.x', 'audit:soc.config.licenseKey'])
self.assertFalse(os.path.exists(soc))
self.assertFalse(os.path.exists(hs))
self.assertIn('action: highstate tgt=*', self.logged('info'))
def test_dispatch_without_jid_not_recorded(self):
self.intent('pillar_soc.json')
with patch.object(drainer, '_dispatch', return_value=''), \
patch.object(drainer, '_record_dispatch') as record, \
patch.object(drainer.os, 'unlink', side_effect=OSError('busy')):
self.assertEqual(drainer.main(), 0)
record.assert_not_called()
self.log.exception.assert_called_once()
def test_untargeted_not_dispatched(self):
path = self.intent('strelka.json', actions=[{'state': 'strelka', 'tgt': 'G@role:so-sensor'}])
self._targeted.side_effect = None
self._targeted.return_value = []
with patch.object(drainer, '_dispatch') as dispatch, \
patch.object(drainer, '_record_dispatch') as record:
self.assertEqual(drainer.main(), 0)
dispatch.assert_not_called()
record.assert_not_called()
self.assertFalse(os.path.exists(path))
self.assertIn('no action matches a minion; nothing to dispatch', self.logged('info'))
def test_dispatches_only_targeted(self):
kept = {'state': 'soc', 'tgt': '*'}
self.intent('mixed.json', actions=[kept, {'state': 'strelka', 'tgt': 'G@role:so-sensor'}])
self._targeted.side_effect = None
self._targeted.return_value = [kept]
with patch.object(drainer, '_dispatch', return_value=JID) as dispatch, \
patch.object(drainer, '_record_dispatch') as record:
self.assertEqual(drainer.main(), 0)
self.assertEqual(dispatch.call_args[0][0], [kept])
self.assertEqual(record.call_args[0][1], [kept])
if __name__ == '__main__':
unittest.main()
+33
View File
@@ -20,6 +20,7 @@ def showUsage(args):
print(' append - Append a list item to a yaml key, if it exists and is a list. Requires KEY and LISTITEM args.', file=sys.stderr)
print(' appendlistobject - Append an object to a yaml list key. Requires KEY and JSON_OBJECT args.', file=sys.stderr)
print(' removelistitem - Remove a list item from a yaml key, if it exists and is a list. Requires KEY and LISTITEM args.', file=sys.stderr)
print(' purgelistitem - Remove every occurrence of a list item from all lists nested under a yaml key, if it exists. Requires KEY and LISTITEM args.', file=sys.stderr)
print(' replacelistobject - Replace a list object based on a condition. Requires KEY, CONDITION_FIELD, CONDITION_VALUE, and JSON_OBJECT args.', file=sys.stderr)
print(' add - Add a new key and set its value. Fails if key already exists. Requires KEY and VALUE args.', file=sys.stderr)
print(' get [-r] - Displays (to stdout) the value stored in the given key. Requires KEY arg. Use -r for raw output without YAML formatting.', file=sys.stderr)
@@ -199,6 +200,33 @@ def removelistitem(args):
return 0
def purgeListItem(content, listItem):
if isinstance(content, dict):
for value in content.values():
purgeListItem(value, listItem)
elif isinstance(content, list):
content[:] = [item for item in content if item != listItem]
for item in content:
purgeListItem(item, listItem)
def purgelistitem(args):
if len(args) != 3:
print('Missing filename, key arg, or list item to purge', file=sys.stderr)
showUsage(None)
return 1
filename = args[0]
key = args[1]
listItem = args[2]
content = loadYaml(filename)
purgeListItem(getKeyValue(content, key), convertType(listItem))
writeYaml(filename, content)
return 0
def replaceListObjectByCondition(content, key, conditionField, conditionValue, newObject):
pieces = key.split(".", 1)
if len(pieces) > 1:
@@ -291,6 +319,8 @@ def add(args):
def removeKey(content, key):
if not isinstance(content, dict):
return
pieces = key.split(".", 1)
if len(pieces) > 1:
if pieces[0] in content:
@@ -334,6 +364,8 @@ def replace(args):
def getKeyValue(content, key):
if not isinstance(content, dict):
return None
pieces = key.split(".", 1)
if len(pieces) > 1 and pieces[0] in content:
return getKeyValue(content[pieces[0]], pieces[1])
@@ -385,6 +417,7 @@ def main():
"append": append,
"appendlistobject": appendlistobject,
"removelistitem": removelistitem,
"purgelistitem": purgelistitem,
"replacelistobject": replacelistobject,
"get": get,
"remove": remove,
+278
View File
@@ -128,6 +128,22 @@ class TestRemove(unittest.TestCase):
sysmock.assert_called_once_with(1)
self.assertIn("Missing filename or key arg\n", mock_stderr.getvalue())
def test_remove_null_parent(self):
filename = "/tmp/so-yaml_test-remove.yaml"
file = open(filename, "w")
file.write("key1: null\n")
file.close()
result = soyaml.remove([filename, "key1.child1.deep1"])
self.assertEqual(result, 0)
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1: null\n"
self.assertEqual(actual, expected)
def test_append_missing_arg(self):
with patch('sys.exit', new=MagicMock()) as sysmock:
with patch('sys.stderr', new=StringIO()) as mock_stderr:
@@ -565,6 +581,19 @@ class TestRemove(unittest.TestCase):
self.assertEqual(result, 2)
self.assertEqual("", mock_stdout.getvalue())
def test_get_null_parent(self):
with patch('sys.stdout', new=StringIO()) as mock_stdout:
with patch('sys.stderr', new=StringIO()) as mock_stderr:
filename = "/tmp/so-yaml_test-get.yaml"
file = open(filename, "w")
file.write("key1: null\n")
file.close()
result = soyaml.get([filename, "key1.child1"])
self.assertEqual(result, 2)
self.assertEqual("", mock_stdout.getvalue())
self.assertIn("Key 'key1.child1' not found by so-yaml.py", mock_stderr.getvalue())
def test_get_empty_file(self):
with patch('sys.stdout', new=StringIO()) as mock_stdout:
with patch('sys.stderr', new=StringIO()) as mock_stderr:
@@ -709,6 +738,255 @@ class TestRemoveListItem(unittest.TestCase):
self.assertEqual("The existing value for the given key is not a list. No action was taken on the file.\n", mock_stderr.getvalue())
class TestPurgeListItem(unittest.TestCase):
def test_purgelistitem_missing_arg(self):
with patch('sys.exit', new=MagicMock()) as sysmock:
with patch('sys.stderr', new=StringIO()) as mock_stderr:
sys.argv = ["cmd", "help"]
soyaml.purgelistitem(["file", "key"])
sysmock.assert_called()
self.assertIn("Missing filename, key arg, or list item to purge", mock_stderr.getvalue())
def test_purgelistitem(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: [a,b,c], key2: false}")
file.close()
soyaml.purgelistitem([filename, "key1", "b"])
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n- a\n- c\nkey2: false\n"
self.assertEqual(actual, expected)
def test_purgelistitem_nested_lists(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: { child1: [a,b,c], child2: { deep1: [b,d], deep2: 45 } }, key2: [b,e]}")
file.close()
soyaml.purgelistitem([filename, "key1", "b"])
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n child1:\n - a\n - c\n child2:\n deep1:\n - d\n deep2: 45\nkey2:\n- b\n- e\n"
self.assertEqual(actual, expected)
def test_purgelistitem_duplicates(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: [b,a,b,c,b]}")
file.close()
soyaml.purgelistitem([filename, "key1", "b"])
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n- a\n- c\n"
self.assertEqual(actual, expected)
def test_purgelistitem_list_of_objects(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: [{name: item1, values: [a,b]}, b, {name: item2, values: [b,c]}]}")
file.close()
soyaml.purgelistitem([filename, "key1", "b"])
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n- name: item1\n values:\n - a\n- name: item2\n values:\n - c\n"
self.assertEqual(actual, expected)
def test_purgelistitem_typed(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: [1, '1', 2]}")
file.close()
soyaml.purgelistitem([filename, "key1", "1"])
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n- '1'\n- 2\n"
self.assertEqual(actual, expected)
def test_purgelistitem_item_not_in_list(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: { child1: [a,b,c] }}")
file.close()
soyaml.purgelistitem([filename, "key1", "d"])
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n child1:\n - a\n - b\n - c\n"
self.assertEqual(actual, expected)
def test_purgelistitem_key_noexist(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: [a,b,c]}")
file.close()
result = soyaml.purgelistitem([filename, "key2.child1", "b"])
self.assertEqual(result, 0)
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n- a\n- b\n- c\n"
self.assertEqual(actual, expected)
def test_purgelistitem_key_noexist_deep(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: { child1: [a,b,c] }}")
file.close()
result = soyaml.purgelistitem([filename, "key1.child2.deep1", "b"])
self.assertEqual(result, 0)
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n child1:\n - a\n - b\n - c\n"
self.assertEqual(actual, expected)
def test_purgelistitem_scalar_and_null_values(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: { child1: b, child2: null, child3: [b] }}")
file.close()
soyaml.purgelistitem([filename, "key1", "b"])
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n child1: b\n child2: null\n child3: []\n"
self.assertEqual(actual, expected)
def test_purgelistitem_lists_of_lists(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: [[a,b], [b,[b,c]], b]}")
file.close()
soyaml.purgelistitem([filename, "key1", "b"])
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n- - a\n- - - c\n"
self.assertEqual(actual, expected)
def test_purgelistitem_file_object(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: [{name: x}, a, {name: y}, {name: x}]}")
file.close()
itemfile = "/tmp/so-yaml_test-purgelistitem-item.yaml"
file = open(itemfile, "w")
file.write("name: x\n")
file.close()
soyaml.purgelistitem([filename, "key1", "file:" + itemfile])
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n- a\n- name: y\n"
self.assertEqual(actual, expected)
def test_purgelistitem_main(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: { child1: [a,b,c] }, key2: [b]}")
file.close()
with patch('sys.exit', new=MagicMock()) as sysmock:
sys.argv = ["cmd", "purgelistitem", filename, "key1", "b"]
soyaml.main()
sysmock.assert_called_once_with(0)
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n child1:\n - a\n - c\nkey2:\n- b\n"
self.assertEqual(actual, expected)
def test_purgelistitem_with_remove(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("{key1: { child1: { deep1: [a] }, child2: [b] }, key2: { group1: { items: [x,child1,y] }, group2: { items: [child1] } }}")
file.close()
soyaml.remove([filename, "key1.child1"])
soyaml.purgelistitem([filename, "key2", "child1"])
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n child2:\n - b\nkey2:\n group1:\n items:\n - x\n - y\n group2:\n items: []\n"
self.assertEqual(actual, expected)
def test_purgelistitem_null_parent(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("key1: null\n")
file.close()
result = soyaml.purgelistitem([filename, "key1.child1", "b"])
self.assertEqual(result, 0)
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1: null\n"
self.assertEqual(actual, expected)
def test_purgelistitem_non_dict_parent(self):
filename = "/tmp/so-yaml_test-purgelistitem.yaml"
file = open(filename, "w")
file.write("key1:\n- a\n- b\n")
file.close()
result = soyaml.purgelistitem([filename, "key1.child1", "b"])
self.assertEqual(result, 0)
file = open(filename, "r")
actual = file.read()
file.close()
expected = "key1:\n- a\n- b\n"
self.assertEqual(actual, expected)
class TestAppendListObject(unittest.TestCase):
def test_appendlistobject_missing_arg(self):
+50 -7
View File
@@ -5,7 +5,6 @@
# https://securityonion.net/license; you may not use this file except in compliance with the
# Elastic License 2.0.
. /usr/sbin/so-common
. /usr/sbin/so-image-common
@@ -17,7 +16,9 @@ INSTALLEDVERSION=$(cat /etc/soversion)
# not finish its post-upgrade steps; its contents are the resume point. It is read
# here before preupgrade_changes mutates INSTALLEDVERSION and before any highstate
# stamps /etc/soversion from the pillar.
POSTVERSION_FILE=/etc/sopostversion
if [ -f "$POSTVERSION_FILE" ]; then
POSTVERSION=$(cat "$POSTVERSION_FILE")
else
@@ -527,6 +528,7 @@ preupgrade_changes() {
[[ "$INSTALLEDVERSION" == "3.1.0" ]] && up_to_3.2.0
[[ "$INSTALLEDVERSION" == "3.2.0" ]] && up_to_3.3.0
[[ "$INSTALLEDVERSION" == "3.3.0" ]] && up_to_3.4.0
[[ "$INSTALLEDVERSION" == "3.4.0" ]] && up_to_3.5.0
true
}
@@ -546,6 +548,7 @@ postupgrade_changes() {
[[ "$POSTVERSION" == "3.1.0" ]] && post_to_3.2.0
[[ "$POSTVERSION" == "3.2.0" ]] && post_to_3.3.0
[[ "$POSTVERSION" == "3.3.0" ]] && post_to_3.4.0
[[ "$POSTVERSION" == "3.4.0" ]] && post_to_3.5.0
# All applicable post-upgrade steps completed; clear the resume marker.
rm -f "$POSTVERSION_FILE"
true
@@ -1065,17 +1068,31 @@ telegraf_repair() {
# when there is something to repair, so everyone else is left alone.
local repair=/usr/sbin/so-telegraf-repair
[[ -x "$repair" ]] || return 0
docker ps --format '{{.Names}}' | grep -qx so-postgres || return 0
local state
state=$(docker inspect -f '{{.State.Running}} {{.State.Restarting}}' so-postgres 2>/dev/null) || return 0
if [[ "$state" != "true false" ]]; then
echo " so-postgres is not running; skipping the Telegraf partition check."
FINAL_MESSAGE_QUEUE+=("WARNING: so-postgres is not running, so Telegraf partitions were not checked. If /nsm is full, follow \"Telegraf Partition Repair\" at:")
FINAL_MESSAGE_QUEUE+=(" - https://securityonion.net/docs/postgresql")
return 0
fi
echo "Checking Telegraf metric partitions."
local status=0
"$repair" --check >> "$SOUP_LOG" 2>&1 || status=$?
local out status=0
out=$("$repair" --check 2>&1) || status=$?
echo "$out" >> "$SOUP_LOG"
case "$status" in
0) echo " Telegraf partitions are healthy; nothing to repair." ;;
1) echo " Repairing stalled Telegraf partitions."
"$repair" --yes \
|| echo " warning: so-telegraf-repair failed; run it manually" >&2 ;;
*) echo " Skipping; Telegraf is not storing metrics in Postgres on this host." ;;
if ! "$repair" --yes; then
echo " warning: so-telegraf-repair failed; run it manually" >&2
FINAL_MESSAGE_QUEUE+=("WARNING: Telegraf partition repair failed; Postgres metrics will keep filling /nsm until it is repaired:")
FINAL_MESSAGE_QUEUE+=(" - sudo so-telegraf-repair")
fi ;;
2) echo " Skipping: $(tail -n 1 <<<"$out")" ;;
*) echo " warning: could not check Telegraf partitions" >&2
FINAL_MESSAGE_QUEUE+=("WARNING: Telegraf partitions could not be checked during the upgrade; run:")
FINAL_MESSAGE_QUEUE+=(" - sudo so-telegraf-repair --check") ;;
esac
}
@@ -1202,6 +1219,32 @@ add_detection_runtime_fields() {
}
### 3.4.0 End ###
### 3.5.0 Scripts ###
up_to_3.5.0() {
remove_localrules_portgroup
INSTALLEDVERSION=3.5.0
}
remove_localrules_portgroup() {
local pillar_file
for pillar_file in /opt/so/saltstack/local/pillar/firewall/{soc,adv}_firewall.sls; do
grep -qs localrules "$pillar_file" || continue
so-yaml.py get "$pillar_file" firewall 2>/dev/null | grep -q '\<localrules\>' || continue
echo "Removing the localrules portgroup from $pillar_file."
so-yaml.py remove "$pillar_file" firewall.portgroups.localrules
so-yaml.py purgelistitem "$pillar_file" firewall.role localrules
done
}
post_to_3.5.0() {
telegraf_repair
set_postversion 3.5.0
}
### 3.5.0 End ###
repo_sync() {
echo "Sync the local repo."
+7 -18
View File
@@ -10,6 +10,13 @@
{% from 'docker/docker.map.jinja' import DOCKERMERGED %}
{% from 'nginx/map.jinja' import NGINXMERGED %}
{% if GLOBALS.role != 'so-fleet' %}
{% set container_config = 'so-nginx' %}
{% else %}
{# if this is an so-fleet node then we want to use the port bindings, custom bind mounts defined for fleet #}
{% set container_config = 'so-nginx-fleet-node' %}
{% endif %}
include:
- ca.trust
- nginx.ssl
@@ -22,23 +29,6 @@ fleet_beats_dir_nginx:
- makedirs: True
- show_changes: False
{% if GLOBALS.role != 'so-fleet' %}
{% set container_config = 'so-nginx' %}
make-rule-dir-nginx:
file.directory:
- name: /nsm/rules
- user: socore
- group: socore
- recurse:
- user
- group
- show_changes: False
{% else %}
{# if this is an so-fleet node then we want to use the port bindings, custom bind mounts defined for fleet #}
{% set container_config = 'so-nginx-fleet-node' %}
{% endif %}
so-nginx:
docker_container.running:
- image: {{ GLOBALS.registry_host }}:5000/{{ GLOBALS.image_repo }}/so-nginx:{{ GLOBALS.so_version }}
@@ -70,7 +60,6 @@ so-nginx:
- /opt/so/conf/navigator/config.json:/opt/socore/html/navigator/assets/config.json:ro
- /nsm/repo:/opt/socore/html/repo:ro
- /nsm/kernelrepo:/opt/socore/html/kernelrepo:ro
- /nsm/rules:/nsm/rules:ro
{% if NGINXMERGED.external_suricata %}
- /opt/so/rules/nids/suri:/surirules:ro
{% endif %}
-15
View File
@@ -92,21 +92,6 @@ http {
{%- endif %}
{%- if GLOBALS.is_manager %}
server {
listen 7788;
server_name {{ full_server_name }};
root /nsm/rules;
location / {
allow all;
sendfile on;
sendfile_max_chunk 1m;
autoindex on;
autoindex_exact_size off;
autoindex_format html;
autoindex_localtime on;
}
}
{%- if NGINXMERGED.external_suricata %}
server {
listen 7789 ssl;
+4 -4
View File
@@ -3,6 +3,8 @@
{% set BATCH = AUTOAPPLY.batch %}
{% set BATCH_WAIT = AUTOAPPLY.batch_wait %}
{# queue must be a top-level salt.state arg (kwarg is ignored); an int is max_queue and still fails on conflict #}
{% for action in actions %}
{% if action.get('highstate') %}
apply_highstate_{{ loop.index }}:
@@ -12,8 +14,7 @@ apply_highstate_{{ loop.index }}:
- highstate: True
- batch: {{ action.get('batch', BATCH) }}
- batch_wait: {{ action.get('batch_wait', BATCH_WAIT) }}
- kwarg:
queue: 2
- queue: True
{% else %}
refresh_pillar_{{ loop.index }}:
salt.function:
@@ -29,8 +30,7 @@ apply_{{ action.state | replace('.', '_') }}_{{ loop.index }}:
- {{ action.state }}
- batch: {{ action.get('batch', BATCH) }}
- batch_wait: {{ action.get('batch_wait', BATCH_WAIT) }}
- kwarg:
queue: 2
- queue: True
- require:
- salt: refresh_pillar_{{ loop.index }}
{% endif %}
+1
View File
@@ -17,3 +17,4 @@ postgres:
log_min_messages: 'warning'
shared_preload_libraries: pg_cron
cron.database_name: postgres
cron.host: /run/postgresql
+6
View File
@@ -87,3 +87,9 @@ postgres:
global: True
advanced: True
helpLink: postgres
cron.host:
description: Where pg_cron connects to run scheduled jobs such as Telegraf partition maintenance. Must be the container's PostgreSQL socket directory, /run/postgresql, because pg_cron connects without a password and only local socket connections are trusted. Any other value makes every scheduled job fail. Changing it restarts PostgreSQL.
readonly: True
global: True
advanced: True
helpLink: postgres
+10 -1
View File
@@ -52,7 +52,7 @@ if [ -n "$FILTER_HOST" ] && ! [[ "$FILTER_HOST" =~ $HOST_RE ]]; then
fi
so_psql() {
docker exec so-postgres psql -U postgres -d so_telegraf -At -F $'\t' "$@"
docker exec -e PGOPTIONS='-c application_name=so-stats-show' so-postgres psql -U so_maint_telegraf -d so_telegraf -At -F $'\t' "$@"
}
if ! docker exec so-postgres psql -U postgres -lqt 2>/dev/null | cut -d\| -f1 | grep -qw so_telegraf; then
@@ -60,6 +60,15 @@ if ! docker exec so-postgres psql -U postgres -lqt 2>/dev/null | cut -d\| -f1 |
exit 2
fi
role_ok=$(docker exec so-postgres psql -U postgres -tAc "SELECT rolcanlogin AND NOT (rolsuper OR rolcreatedb OR rolcreaterole OR rolreplication OR rolbypassrls) FROM pg_roles WHERE rolname = 'so_maint_telegraf'" 2>/dev/null)
if [ -z "$role_ok" ]; then
echo "The so_maint_telegraf role is missing; run a highstate."
exit 1
elif [ "$role_ok" != t ]; then
echo "The so_maint_telegraf role is misconfigured; run a highstate."
exit 1
fi
table_exists() {
local table="$1"
[ -n "$(so_psql -c "SELECT 1 FROM information_schema.tables WHERE table_schema='${SCHEMA}' AND table_name='${table}' LIMIT 1;")" ]
+196 -70
View File
@@ -4,15 +4,18 @@ set -e
# Provision Telegraf state inside the so-postgres container.
# Usage: so-telegraf-postgres <subcommand>
# create_db Ensure the so_telegraf database exists.
# group_role Provision the so_telegraf group role, telegraf/partman schemas,
# pg_partman, the so_admin maintenance routines, and the hourly
# pg_cron maintenance job.
# group_role Provision the so_telegraf group role, the so_maint_telegraf
# maintenance login role, telegraf/partman schemas, pg_partman,
# the so_admin maintenance routines, and the hourly pg_cron
# maintenance job.
# user Create or update a per-minion login role granted to so_telegraf.
# Env: ROLE_USER, ROLE_PASS.
# retention Reconcile partman retention and premake on telegraf parents.
# Env: RETENTION_DAYS.
# maintenance Drain default partitions and run partman maintenance.
# check Report partition health. Non-zero if any parent is unhealthy.
# maintenance Drain default partitions and run partman maintenance, as
# so_maint_telegraf.
# check Report partition health, as so_maint_telegraf. Non-zero if any
# parent is unhealthy.
#
# A default partition holding rows for a day blocks creating that day's child,
# so maintenance drains defaults before calling partman. Use so-telegraf-repair
@@ -20,6 +23,10 @@ set -e
cmd="${1:?subcommand required}"
# Long enough for an hourly maintenance run to release its locks, short enough
# that soup and admins are not left hanging.
LOCK_TIMEOUT=60s
case "$cmd" in
create_db)
if ! docker exec so-postgres psql -U postgres -tAc \
@@ -30,7 +37,9 @@ case "$cmd" in
;;
group_role)
docker exec -i so-postgres psql -v ON_ERROR_STOP=1 -U postgres -d so_telegraf <<'EOSQL'
docker exec -i so-postgres psql -v ON_ERROR_STOP=1 -v lock_timeout="$LOCK_TIMEOUT" \
-U postgres -d so_telegraf <<'EOSQL'
SET lock_timeout = :'lock_timeout';
DO $$
BEGIN
IF NOT EXISTS (SELECT FROM pg_catalog.pg_roles WHERE rolname = 'so_telegraf') THEN
@@ -38,6 +47,17 @@ BEGIN
END IF;
END
$$;
DO $$
BEGIN
IF NOT EXISTS (SELECT FROM pg_catalog.pg_roles WHERE rolname = 'so_maint_telegraf') THEN
CREATE ROLE so_maint_telegraf;
END IF;
END
$$;
ALTER ROLE so_maint_telegraf LOGIN NOSUPERUSER NOCREATEDB NOCREATEROLE NOREPLICATION NOBYPASSRLS PASSWORD NULL;
GRANT so_telegraf TO so_maint_telegraf;
ALTER ROLE so_maint_telegraf SET role = so_telegraf;
ALTER ROLE so_maint_telegraf SET search_path = pg_catalog;
GRANT CONNECT ON DATABASE so_telegraf TO so_telegraf;
CREATE SCHEMA IF NOT EXISTS telegraf AUTHORIZATION so_telegraf;
GRANT USAGE, CREATE ON SCHEMA telegraf TO so_telegraf;
@@ -58,52 +78,90 @@ ALTER DEFAULT PRIVILEGES IN SCHEMA partman
ALTER DEFAULT PRIVILEGES IN SCHEMA partman
GRANT USAGE, SELECT, UPDATE ON SEQUENCES TO so_telegraf;
-- pg_cron runs these as postgres, so they must not sit in a schema any
-- Telegraf role can create objects in.
-- Owned by postgres so no Telegraf role can create objects here; the routines
-- run with the caller's privileges.
CREATE SCHEMA IF NOT EXISTS so_admin AUTHORIZATION postgres;
REVOKE ALL ON SCHEMA so_admin FROM PUBLIC;
-- Resolved through pg_inherits, never by name, so a view or a table that only
-- carries the _default name is never read or written.
CREATE OR REPLACE FUNCTION so_admin.telegraf_defaults()
RETURNS TABLE (
parent_table regclass,
default_part regclass,
has_rows boolean,
default_size bigint
)
LANGUAGE plpgsql
STABLE
SECURITY INVOKER
AS $func$
DECLARE
r record;
BEGIN
FOR r IN
SELECT c.oid::regclass AS pt
FROM partman.part_config pc
JOIN pg_catalog.pg_class c ON c.oid = to_regclass(pc.parent_table)
WHERE pc.parent_table LIKE 'telegraf.%'
AND c.relkind = 'p'
ORDER BY pc.parent_table
LOOP
parent_table := r.pt;
SELECT c.oid::regclass INTO default_part
FROM pg_catalog.pg_inherits i
JOIN pg_catalog.pg_class c ON c.oid = i.inhrelid
WHERE i.inhparent = r.pt
AND c.relkind = 'r' AND c.relispartition
AND pg_catalog.pg_get_expr(c.relpartbound, c.oid) = 'DEFAULT';
IF default_part IS NULL THEN
has_rows := false;
default_size := NULL;
ELSE
EXECUTE format('SELECT EXISTS (SELECT 1 FROM %s LIMIT 1)', default_part)
INTO has_rows;
default_size := pg_catalog.pg_total_relation_size(default_part);
END IF;
RETURN NEXT;
END LOOP;
END;
$func$;
CREATE OR REPLACE PROCEDURE so_admin.telegraf_maintenance()
LANGUAGE plpgsql
SECURITY INVOKER
-- No SET clause: it would forbid the COMMITs below, so so_maint_telegraf pins search_path instead.
AS $proc$
DECLARE
r record;
v_default text;
v_rows bigint;
r record;
BEGIN
-- No per-parent EXCEPTION handler: partition_data_proc commits internally,
-- and COMMIT is illegal while a subtransaction is active. A failing parent
-- aborts the run and the next pass retries.
FOR r IN
SELECT parent_table, retention
FROM partman.part_config
WHERE parent_table LIKE 'telegraf.%'
ORDER BY parent_table
SELECT pc.parent_table, pc.retention, d.default_part, d.default_size
FROM so_admin.telegraf_defaults() d
JOIN partman.part_config pc ON to_regclass(pc.parent_table) = d.parent_table
WHERE d.default_part IS NOT NULL AND d.has_rows
ORDER BY pc.parent_table
LOOP
v_default := format('%I.%I',
split_part(r.parent_table, '.', 1),
split_part(r.parent_table, '.', 2) || '_default');
CONTINUE WHEN to_regclass(v_default) IS NULL;
EXECUTE format('SELECT count(*) FROM %s', v_default) INTO v_rows;
CONTINUE WHEN v_rows = 0;
RAISE WARNING 'so_admin.telegraf_maintenance: % rows stranded in %, draining',
v_rows, v_default;
RAISE WARNING 'so_admin.telegraf_maintenance: % holds rows (%), draining',
r.default_part, pg_size_pretty(r.default_size);
-- Cheaper to delete expired rows than to repartition and then drop them.
IF r.retention IS NOT NULL THEN
EXECUTE format('DELETE FROM %s WHERE "time" < now() - %L::interval',
v_default, r.retention);
r.default_part, r.retention);
COMMIT;
END IF;
-- Bounded so a large backlog drains across several runs.
CALL partman.partition_data_proc(
p_parent_table := r.parent_table,
p_loop_count := 200,
p_source_table := v_default
p_loop_count := 200
);
COMMIT;
END LOOP;
@@ -112,58 +170,75 @@ BEGIN
END;
$proc$;
CREATE OR REPLACE FUNCTION so_admin.telegraf_partition_status()
DROP FUNCTION IF EXISTS so_admin.telegraf_partition_status();
CREATE FUNCTION so_admin.telegraf_partition_status()
RETURNS TABLE (
parent_table text,
oldest_child date,
newest_child date,
days_ahead int,
retention text,
default_rows bigint,
default_size text
parent_table text,
oldest_child date,
newest_child date,
days_ahead int,
retention text,
default_has_rows boolean,
default_size text
)
LANGUAGE plpgsql
STABLE
AS $func$
DECLARE
r record;
v_default regclass;
r record;
BEGIN
FOR r IN
SELECT pc.parent_table AS pt, pc.retention AS ret
FROM partman.part_config pc
WHERE pc.parent_table LIKE 'telegraf.%'
ORDER BY pc.parent_table
SELECT d.parent_table AS pt, d.has_rows, d.default_size AS size, pc.retention AS ret
FROM so_admin.telegraf_defaults() d
JOIN partman.part_config pc ON to_regclass(pc.parent_table) = d.parent_table
ORDER BY d.parent_table::text
LOOP
parent_table := r.pt;
parent_table := r.pt::text;
retention := r.ret;
SELECT min(d), max(d) INTO oldest_child, newest_child
SELECT min(x), max(x) INTO oldest_child, newest_child
FROM (
SELECT to_date(substring(c.relname FROM '_p(\d{8})$'), 'YYYYMMDD') AS d
FROM pg_inherits i
JOIN pg_class c ON c.oid = i.inhrelid
WHERE i.inhparent = r.pt::regclass
AND pg_get_expr(c.relpartbound, c.oid) <> 'DEFAULT'
SELECT to_date(substring(c.relname FROM '_p(\d{8})$'), 'YYYYMMDD') AS x
FROM pg_catalog.pg_inherits i
JOIN pg_catalog.pg_class c ON c.oid = i.inhrelid
WHERE i.inhparent = r.pt
AND pg_catalog.pg_get_expr(c.relpartbound, c.oid) <> 'DEFAULT'
) s;
days_ahead := newest_child - current_date;
v_default := to_regclass(format('%I.%I',
split_part(r.pt, '.', 1),
split_part(r.pt, '.', 2) || '_default'));
IF v_default IS NULL THEN
default_rows := NULL;
default_size := NULL;
ELSE
EXECUTE format('SELECT count(*) FROM %s', v_default::text) INTO default_rows;
default_size := pg_size_pretty(pg_total_relation_size(v_default));
END IF;
days_ahead := newest_child - current_date;
default_has_rows := r.has_rows;
default_size := pg_size_pretty(r.size);
RETURN NEXT;
END LOOP;
END;
$func$;
GRANT USAGE ON SCHEMA so_admin TO so_telegraf;
GRANT EXECUTE ON FUNCTION so_admin.telegraf_defaults() TO so_telegraf;
GRANT EXECUTE ON PROCEDURE so_admin.telegraf_maintenance() TO so_telegraf;
GRANT EXECUTE ON FUNCTION so_admin.telegraf_partition_status() TO so_telegraf;
-- Partitions created by a superuser session (an older repair) belong to
-- postgres; partman acting as so_telegraf could not drop them.
DO $$
DECLARE
r record;
BEGIN
FOR r IN
SELECT n.nspname, c.relname
FROM pg_catalog.pg_class c
JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace
WHERE c.relkind IN ('r', 'p')
AND (n.nspname = 'telegraf'
OR (n.nspname = 'partman' AND c.relname LIKE 'template\_telegraf\_%'))
AND c.relowner <> 'so_telegraf'::regrole
LOOP
EXECUTE format('ALTER TABLE %I.%I OWNER TO so_telegraf', r.nspname, r.relname);
END LOOP;
END
$$;
-- Drop the registration older releases left in so_telegraf.
SELECT CASE
WHEN current_setting('cron.database_name', true) IS DISTINCT FROM current_database()
@@ -175,20 +250,69 @@ DROP EXTENSION pg_cron CASCADE;
\endif
EOSQL
stalled=false
probe_failed=false
if out=$(docker exec -e PGOPTIONS="-c lock_timeout=$LOCK_TIMEOUT" so-postgres \
psql -U so_maint_telegraf -d so_telegraf -tAc \
"SELECT coalesce(bool_or(has_rows), false) FROM so_admin.telegraf_defaults()"); then
case "$out" in
t) stalled=true ;;
f) stalled=false ;;
*) probe_failed=true ;;
esac
else
probe_failed=true
fi
if $probe_failed; then
echo "ERROR: could not check Telegraf default partitions; the maintenance job was not changed." >&2
exit 1
fi
# Guarded on the live GUC so applying this before the postgresql.conf change
# has restarted the container skips rather than failing.
docker exec -i so-postgres psql -v ON_ERROR_STOP=1 -U postgres -d postgres <<'EOSQL'
docker exec -i so-postgres psql -v ON_ERROR_STOP=1 -v stalled="$stalled" \
-U postgres -d postgres <<'EOSQL'
SELECT CASE WHEN current_setting('cron.database_name', true) = current_database()
THEN 'true' ELSE 'false' END AS cron_here \gset
\if :cron_here
CREATE EXTENSION IF NOT EXISTS pg_cron;
-- cron.schedule_in_database is idempotent by jobname.
-- A highstate reconciles the job's definition but never its active flag; only so-telegraf-repair turns it on.
BEGIN;
SELECT coalesce(max(jobid)::text, '') AS jobid,
count(*) > 0 AS job_exists,
coalesce(bool_and(active), true) AS job_active
FROM cron.job
WHERE jobname = 'telegraf-partman-maintenance' AND username = 'so_maint_telegraf' \gset
SELECT cron.unschedule(jobid) FROM cron.job
WHERE jobname = 'telegraf-partman-maintenance' AND username <> 'so_maint_telegraf';
\if :job_exists
SELECT cron.alter_job(:jobid,
schedule := '17 * * * *',
command := 'CALL so_admin.telegraf_maintenance()',
database := 'so_telegraf');
\else
-- Registered inactive on a stalled grid so nothing drains the default
-- partitions before so-telegraf-repair truncates them.
SELECT cron.schedule_in_database(
'telegraf-partman-maintenance',
'17 * * * *',
'CALL so_admin.telegraf_maintenance()',
'so_telegraf'
'so_telegraf',
'so_maint_telegraf',
NOT :'stalled'::boolean
);
\endif
UPDATE cron.job SET nodename = current_setting('cron.host')
WHERE jobname = 'telegraf-partman-maintenance'
AND nodename IS DISTINCT FROM current_setting('cron.host');
COMMIT;
SELECT coalesce(bool_and(active), true) AS job_active
FROM cron.job
WHERE jobname = 'telegraf-partman-maintenance' AND username = 'so_maint_telegraf' \gset
\if :job_active
\else
\echo 'WARNING: the Telegraf maintenance job is inactive; run ''sudo so-telegraf-repair''.'
\endif
\else
\echo 'pg_cron metadata database is not `postgres` yet; skipping job registration.'
\endif
@@ -241,13 +365,15 @@ SET retention = :'retention_days' || ' days',
retention_keep_table = false,
premake = 7,
infinite_time_partitions = true
WHERE parent_table LIKE 'telegraf.%';
WHERE parent_table LIKE 'telegraf.%'
AND EXISTS (SELECT 1 FROM pg_catalog.pg_class c
WHERE c.oid = to_regclass(parent_table) AND c.relkind = 'p');
\endif
EOSQL
;;
maintenance)
docker exec -i so-postgres psql -v ON_ERROR_STOP=1 -U postgres -d so_telegraf <<'EOSQL'
docker exec -i so-postgres psql -v ON_ERROR_STOP=1 -U so_maint_telegraf -d so_telegraf <<'EOSQL'
SELECT CASE WHEN to_regproc('so_admin.telegraf_maintenance') IS NOT NULL
THEN 'true' ELSE 'false' END AS has_proc \gset
\if :has_proc
@@ -259,7 +385,7 @@ EOSQL
;;
check)
docker exec -i so-postgres psql -U postgres -d so_telegraf <<'EOSQL'
docker exec -i so-postgres psql -U so_maint_telegraf -d so_telegraf <<'EOSQL'
\pset border 2
SELECT * FROM so_admin.telegraf_partition_status();
EOSQL
@@ -278,9 +404,9 @@ LIMIT 5;
\echo 'pg_cron is not installed in this database.'
\endif
EOSQL
unhealthy=$(docker exec so-postgres psql -U postgres -d so_telegraf -tAc \
unhealthy=$(docker exec so-postgres psql -U so_maint_telegraf -d so_telegraf -tAc \
"SELECT count(*) FROM so_admin.telegraf_partition_status()
WHERE coalesce(default_rows, 0) > 0 OR coalesce(days_ahead, -1) < 1")
WHERE default_has_rows OR coalesce(days_ahead, -1) < 1")
if [ "${unhealthy:-1}" != "0" ]; then
echo "so-telegraf-postgres check: $unhealthy telegraf parent(s) unhealthy" >&2
exit 1
+225 -83
View File
@@ -11,26 +11,38 @@
# maintenance once. Healthy grids are reported and left alone.
#
# Usage: so-telegraf-repair [--check] [--yes] [--no-restart]
# --check Report health and change nothing.
# --check Report health and change nothing. Also reports a missing
# so_maint_telegraf role and a maintenance job that is missing
# or inactive.
# --yes Skip the confirmation prompt (for soup and other automation).
# --no-restart Never restart so-postgres, even if pg_cron's launcher is dead.
#
# Repair mode (without --check) registers the so_maint_telegraf role and the
# maintenance job if they are missing, pauses the job while it repairs, and
# always leaves it active afterwards. Data work runs as so_maint_telegraf.
#
# Exit status:
# 0 healthy, or repair completed
# 1 repair is needed (--check only)
# 2 cannot run here: so-postgres, so_telegraf or pg_partman is missing
# 1 repair is needed (--check), or the repair failed after it started
# 2 skipped: Telegraf output is not Postgres, or so-postgres, so_telegraf or pg_partman is missing
# 3 could not check, for example a connection error
set -e
set -eE
# Matches p_premake in telegraf.conf's create_parent template.
PREMAKE=7
JOB_NAME=telegraf-partman-maintenance
MAINT_ROLE=so_maint_telegraf
# Long enough for an hourly maintenance run to release its locks, short enough
# that soup and admins are not left hanging.
LOCK_TIMEOUT=60s
PG_OPTS="-c lock_timeout=$LOCK_TIMEOUT -c application_name=so-telegraf-repair"
CHECK_ONLY=false
ASSUME_YES=false
NO_RESTART=false
usage() { sed -n '/^# Usage:/,/^# 2 /p' "$0" | sed 's/^# \?//'; }
usage() { sed -n '/^# Usage:/,/^# 3 /p' "$0" | sed 's/^# \?//'; }
while [[ $# -gt 0 ]]; do
case "$1" in
@@ -38,75 +50,170 @@ while [[ $# -gt 0 ]]; do
--yes|-y) ASSUME_YES=true ;;
--no-restart) NO_RESTART=true ;;
-h|--help) usage; exit 0 ;;
*) echo "Unknown option: $1" >&2; usage >&2; exit 2 ;;
*) echo "Unknown option: $1" >&2; usage >&2; exit 3 ;;
esac
shift
done
skip() { echo "$*"; exit 2; }
# Strips C0 (except tab and newline), DEL and C1 characters, so names and
# messages written with Telegraf credentials can't drive the terminal.
sanitise() { LC_ALL=C tr -d '\000-\010\013-\037\177' | LC_ALL=C sed 's/\xc2[\x80-\x9f]//g'; }
psql_tg() { docker exec -i so-postgres psql -U postgres -d so_telegraf "$@"; }
psql_pg() { docker exec -i so-postgres psql -U postgres -d postgres "$@"; }
ERR_FILE=$(mktemp)
ERR_STATUS=3
ERR_PREFIX="Could not check Telegraf partitions"
job_id=
restore_active=
on_err() {
# Command substitutions inherit this trap; only the top level reports.
[[ $BASH_SUBSHELL -eq 0 ]] || exit "$ERR_STATUS"
local line
line=$(tail -n 1 "$ERR_FILE" | sanitise)
echo "$ERR_PREFIX: ${line:-command failed}" >&2
exit "$ERR_STATUS"
}
on_exit() {
if [[ -n "$restore_active" ]]; then
psql_cron -tAc "SELECT cron.alter_job($job_id, active := $restore_active)" >/dev/null || true
fi
rm -f "$ERR_FILE"
}
trap on_err ERR
trap on_exit EXIT
finish() { trap - ERR; exit "$1"; }
skip() { echo "$*"; finish 2; }
# Keeps the last stderr line for on_err while still showing all of it.
capture() {
local rc=0
: >"$ERR_FILE"
"$@" 2>"$ERR_FILE" || rc=$?
sanitise <"$ERR_FILE" >&2
return "$rc"
}
psql_tg() {
capture docker exec -i -e PGOPTIONS="$PG_OPTS" so-postgres psql -U "$MAINT_ROLE" -d so_telegraf "$@"
}
psql_pg() {
capture docker exec -i -e PGOPTIONS="$PG_OPTS" so-postgres psql -U postgres -d postgres "$@"
}
psql_cron() {
capture docker exec -i -e PGOPTIONS="$PG_OPTS" so-postgres psql -U postgres -d "$cron_db" "$@"
}
# A part_config row counts only while its parent is a partitioned table.
TELEGRAF_PARENTS="parent_table LIKE 'telegraf.%'
AND EXISTS (SELECT 1 FROM pg_class c
WHERE c.oid = to_regclass(parent_table) AND c.relkind = 'p')"
# query_to_xml so the per-table row counts need no helper function installed.
REPORT="
WITH parents AS (
SELECT pc.parent_table,
pc.premake,
split_part(pc.parent_table, '.', 1) AS sch,
split_part(pc.parent_table, '.', 2) AS tbl
FROM partman.part_config pc
WHERE pc.parent_table LIKE 'telegraf.%'
), children AS (
SELECT p.parent_table,
max(to_date(substring(c.relname FROM '_p(\d{8})\$'), 'YYYYMMDD')) AS newest_child
FROM parents p
JOIN pg_class pt ON pt.oid = p.parent_table::regclass
JOIN pg_inherits i ON i.inhparent = pt.oid
JOIN pg_class c ON c.oid = i.inhrelid
WHERE pg_get_expr(c.relpartbound, c.oid) <> 'DEFAULT'
GROUP BY p.parent_table
), defaults AS (
SELECT p.parent_table,
p.premake,
format('%I.%I', p.sch, p.tbl || '_default') AS default_table,
to_regclass(format('%I.%I', p.sch, p.tbl || '_default')) AS default_oid
FROM parents p
)
SELECT d.parent_table,
c.newest_child,
(c.newest_child - current_date) AS days_ahead,
d.premake,
CASE WHEN d.default_oid IS NULL THEN NULL ELSE
(xpath('/row/cnt/text()',
query_to_xml(format('SELECT count(*) AS cnt FROM %s', d.default_table),
false, true, '')))[1]::text::bigint
END AS default_rows,
CASE WHEN d.default_oid IS NULL THEN NULL
ELSE pg_size_pretty(pg_total_relation_size(d.default_oid)) END AS default_size
FROM defaults d
LEFT JOIN children c ON c.parent_table = d.parent_table
ORDER BY 1
pc.premake,
d.has_rows AS default_has_rows,
pg_size_pretty(d.default_size) AS default_size
FROM so_admin.telegraf_defaults() d
JOIN partman.part_config pc ON to_regclass(pc.parent_table) = d.parent_table
JOIN pg_class p ON p.oid = d.parent_table AND p.relkind = 'p'
LEFT JOIN LATERAL (
SELECT max(to_date(substring(ch.relname FROM '_p(\d{8})\$'), 'YYYYMMDD')) AS newest_child
FROM pg_inherits i
JOIN pg_class ch ON ch.oid = i.inhrelid
WHERE i.inhparent = d.parent_table
AND pg_get_expr(ch.relpartbound, ch.oid) <> 'DEFAULT'
) c ON true
ORDER BY d.parent_table::text
"
docker ps --format '{{.Names}}' | grep -qx so-postgres \
print_report() {
local report
report=$(psql_tg -c "$REPORT")
sanitise <<<"$report"
}
has_cron() {
[[ -n "$cron_db" ]] || return 0
psql_cron -tAc "SELECT 1 FROM pg_extension WHERE extname = 'pg_cron'"
}
lookup_job() {
local cron
cron=$(has_cron)
[[ -n "$cron" ]] || return 0
psql_cron -tA -F '|' -c \
"SELECT jobid, active FROM cron.job
WHERE jobname = '$JOB_NAME' AND username = '$MAINT_ROLE'
ORDER BY jobid LIMIT 1"
}
output=$(capture salt-call pillar.get telegraf:output --out=newline_values_only)
output=$(tr -d '[:space:]' <<<"$output" | tr '[:lower:]' '[:upper:]')
output=${output:-INFLUXDB}
case "$output" in
POSTGRES|BOTH) ;;
*) skip "Telegraf output is $(sanitise <<<"$output"); not checking Postgres partitions." ;;
esac
containers=$(capture docker ps --format '{{.Names}}')
grep -qx so-postgres <<<"$containers" \
|| skip "so-postgres is not running; nothing to repair."
docker exec so-postgres psql -U postgres -tAc \
"SELECT 1 FROM pg_database WHERE datname='so_telegraf'" | grep -q 1 \
has_db=$(psql_pg -tAc "SELECT 1 FROM pg_database WHERE datname = 'so_telegraf'")
[[ -n "$has_db" ]] \
|| skip "The so_telegraf database does not exist; Telegraf is not writing to Postgres."
psql_tg -tAc "SELECT 1 FROM pg_extension WHERE extname='pg_partman'" | grep -q 1 \
# so_telegraf before the postgres state lands, postgres after.
cron_db=$(psql_pg -tAc "SELECT current_setting('cron.database_name', true)")
cron_db=$(tr -d '[:space:]' <<<"$cron_db")
role_ok=$(psql_pg -tAc \
"SELECT rolcanlogin AND NOT (rolsuper OR rolcreatedb OR rolcreaterole OR rolreplication OR rolbypassrls)
FROM pg_roles WHERE rolname = '$MAINT_ROLE'")
job=
[[ "$cron_db" == postgres ]] && job=$(lookup_job)
if [[ "$role_ok" != t ]] && $CHECK_ONLY; then
echo "Repair is needed:"
if [[ -z "$role_ok" ]]; then
echo " * maintenance role $MAINT_ROLE is missing"
else
echo " * maintenance role $MAINT_ROLE is misconfigured"
fi
echo
echo "Re-run without --check to repair."
finish 1
fi
if ! $CHECK_ONLY && [[ "$role_ok" != t || ( "$cron_db" == postgres && -z "$job" ) ]]; then
echo "Registering the maintenance role and job."
if ! /usr/sbin/so-telegraf-postgres group_role; then
echo "Could not register the maintenance job; run a highstate and try again." >&2
finish 3
fi
[[ "$cron_db" == postgres ]] && job=$(lookup_job)
fi
has_partman=$(psql_tg -tAc "SELECT 1 FROM pg_extension WHERE extname = 'pg_partman'")
[[ -n "$has_partman" ]] \
|| skip "pg_partman is not installed in so_telegraf; nothing to repair."
parents=$(psql_tg -tAc \
"SELECT count(*) FROM partman.part_config WHERE parent_table LIKE 'telegraf.%'")
stranded=$(psql_tg -tAc "SELECT coalesce(sum(default_rows), 0) FROM ( $REPORT ) t")
behind=$(psql_tg -tAc \
"SELECT count(*) FROM ( $REPORT ) t WHERE coalesce(days_ahead, -1) < 1")
"SELECT count(*) FROM partman.part_config WHERE $TELEGRAF_PARENTS")
counts=$(psql_tg -tAc \
"SELECT count(*) FILTER (WHERE default_has_rows) || '|' ||
count(*) FILTER (WHERE coalesce(days_ahead, -1) < 1)
FROM ( $REPORT ) t")
IFS='|' read -r stranded behind <<<"$counts"
# premake < 7, or infinite_time_partitions off: without the latter partman
# refuses to premake forward across the gap the stall left behind.
misconfigured=$(psql_tg -tAc \
"SELECT count(*) FROM partman.part_config
WHERE parent_table LIKE 'telegraf.%'
WHERE $TELEGRAF_PARENTS
AND (premake < $PREMAKE OR NOT infinite_time_partitions)")
# Both columns are matched because which one carries the launcher's name varies
@@ -115,16 +222,24 @@ launcher=$(psql_pg -tAc \
"SELECT count(*) FROM pg_stat_activity
WHERE backend_type ILIKE '%pg_cron%' OR application_name ILIKE '%pg_cron%'")
# so_telegraf before the postgres state lands, postgres after.
cron_db=$(docker exec so-postgres psql -U postgres -tAc \
"SELECT current_setting('cron.database_name', true)" | tr -d '[:space:]')
last_run=never
if [[ -n "$cron_db" ]]; then
last_run=$(docker exec so-postgres psql -U postgres -d "$cron_db" -tAc \
"SELECT coalesce(max(d.start_time)::text, 'never')
cron_installed=$(has_cron)
if [[ -n "$cron_installed" ]]; then
run=$(psql_cron -tAc \
"SELECT d.status || ' at ' || d.start_time
FROM cron.job_run_details d JOIN cron.job j USING (jobid)
WHERE j.jobname = '$JOB_NAME'" 2>/dev/null | tr -d '[:space:]' || echo unknown)
[[ -n "$last_run" ]] || last_run=never
WHERE j.jobname = '$JOB_NAME'
ORDER BY d.start_time DESC LIMIT 1")
if [[ -n "$run" ]]; then
message=$(psql_cron -tAc \
"SELECT coalesce(d.return_message, '')
FROM cron.job_run_details d JOIN cron.job j USING (jobid)
WHERE j.jobname = '$JOB_NAME'
ORDER BY d.start_time DESC LIMIT 1")
message=$(sanitise <<<"$message")
message=${message//$'\n'/ }
last_run="$(sanitise <<<"$run"): ${message:0:200}"
fi
fi
# A grid that has never written a metric has nothing to recover, and an empty
@@ -132,43 +247,67 @@ fi
restart_needed=false
[[ "$launcher" -eq 0 && "$parents" -gt 0 && -n "$cron_db" ]] && restart_needed=true
# group_role registers the job only in the postgres database.
job_inactive=false
if [[ "$cron_db" == postgres ]]; then
[[ -z "$job" || "${job##*|}" == f ]] && job_inactive=true
fi
repair_needed=false
[[ "$stranded" -gt 0 ]] && repair_needed=true
[[ "$behind" -gt 0 ]] && repair_needed=true
[[ "$misconfigured" -gt 0 ]] && repair_needed=true
$restart_needed && repair_needed=true
$job_inactive && repair_needed=true
echo "Telegraf partition status:"
psql_tg -c "$REPORT"
echo "Rows stranded in default partitions: $stranded"
echo "pg_cron metadata database: ${cron_db:-unset}"
print_report
echo "Default partitions with rows: $stranded"
echo "pg_cron metadata database: $(sanitise <<<"${cron_db:-unset}")"
[[ "$cron_db" == postgres ]] \
|| echo "pg_cron metadata database is $(sanitise <<<"${cron_db:-unset}"); not checking the maintenance job."
echo "pg_cron launcher running: $([[ "$launcher" -gt 0 ]] && echo yes || echo no)"
echo "Last $JOB_NAME run: $last_run"
echo
if ! $repair_needed; then
echo "Telegraf partitions are healthy. Nothing to do."
exit 0
finish 0
fi
if $CHECK_ONLY; then
echo "Repair is needed:"
[[ "$stranded" -gt 0 ]] && echo " * $stranded row(s) stranded in default partitions"
[[ "$stranded" -gt 0 ]] && echo " * $stranded default partition(s) hold rows"
[[ "$behind" -gt 0 ]] && echo " * $behind parent(s) with no partition for the current window"
[[ "$misconfigured" -gt 0 ]] && echo " * $misconfigured parent(s) with stale partman settings"
$restart_needed && echo " * pg_cron's launcher is dead; maintenance is not running at all"
$job_inactive && echo " * the maintenance job is missing or inactive"
echo
echo "Re-run without --check to repair."
exit 1
finish 1
fi
if [[ "$stranded" -gt 0 ]] && ! $ASSUME_YES; then
echo "This will permanently discard the $stranded stranded row(s) above."
echo "This will permanently discard the rows in the $stranded default partition(s) above."
$restart_needed && ! $NO_RESTART && \
echo "so-postgres will also be restarted, which briefly interrupts SOC."
[[ -t 0 ]] || { echo "Not a terminal; re-run with --yes to confirm." >&2; exit 2; }
[[ -t 0 ]] || { echo "Not a terminal; re-run with --yes to confirm." >&2; finish 1; }
read -r -p "Continue? [y/N] " answer
[[ "$answer" =~ ^[Yy]$ ]] || { echo "Aborted."; exit 0; }
[[ "$answer" =~ ^[Yy]$ ]] || { echo "Aborted."; finish 0; }
fi
ERR_STATUS=1
ERR_PREFIX="Telegraf partition repair failed"
if [[ "$cron_db" == postgres ]]; then
if [[ -z "$job" ]]; then
echo "Could not register the maintenance job; run a highstate and try again." >&2
finish 3
fi
job_id=${job%%|*}
[[ "${job##*|}" == t ]] && restore_active=true || restore_active=false
echo "Pausing the $JOB_NAME job during the repair."
psql_cron -tAc "SELECT cron.alter_job($job_id, active := false)" >/dev/null
fi
if [[ "$misconfigured" -gt 0 ]]; then
@@ -178,7 +317,7 @@ if [[ "$misconfigured" -gt 0 ]]; then
"UPDATE partman.part_config
SET premake = GREATEST(premake, $PREMAKE),
infinite_time_partitions = true
WHERE parent_table LIKE 'telegraf.%'"
WHERE $TELEGRAF_PARENTS"
fi
if [[ "$stranded" -gt 0 ]]; then
@@ -186,27 +325,24 @@ if [[ "$stranded" -gt 0 ]]; then
# One transaction: Telegraf is still writing, so a default emptied without
# its partition in place is refilled before maintenance can attach one.
psql_tg -v ON_ERROR_STOP=1 <<'EOSQL'
BEGIN;
DO $$
DECLARE
r record;
BEGIN
FOR r IN
SELECT pc.parent_table,
format('%I.%I', n.nspname, c.relname) AS default_table
FROM partman.part_config pc
JOIN pg_class p ON p.oid = pc.parent_table::regclass
JOIN pg_inherits i ON i.inhparent = p.oid
JOIN pg_class c ON c.oid = i.inhrelid
JOIN pg_namespace n ON n.oid = c.relnamespace
WHERE pc.parent_table LIKE 'telegraf.%'
AND pg_get_expr(c.relpartbound, c.oid) = 'DEFAULT'
SELECT pc.parent_table, d.default_part
FROM so_admin.telegraf_defaults() d
JOIN partman.part_config pc ON to_regclass(pc.parent_table) = d.parent_table
WHERE d.default_part IS NOT NULL
LOOP
EXECUTE format('TRUNCATE TABLE %s', r.default_table);
EXECUTE format('TRUNCATE TABLE %s', r.default_part);
PERFORM partman.create_partition_time(
r.parent_table, ARRAY[date_trunc('day', now())]::timestamptz[]);
END LOOP;
END
$$;
COMMIT;
EOSQL
fi
@@ -222,7 +358,7 @@ if $restart_needed; then
sleep 2
done
docker exec so-postgres pg_isready -U postgres -q \
|| { echo "so-postgres did not come back; check 'docker logs so-postgres'." >&2; exit 1; }
|| { echo "so-postgres did not come back; check 'docker logs so-postgres'." >&2; finish 1; }
fi
fi
@@ -238,8 +374,14 @@ CALL partman.run_maintenance_proc();
\endif
EOSQL
if [[ -n "$job_id" ]]; then
psql_cron -tAc "SELECT cron.alter_job($job_id, active := true)" >/dev/null
restore_active=
echo "The $JOB_NAME job is active."
fi
echo
echo "Telegraf partition status after repair:"
psql_tg -c "$REPORT"
print_report
echo "The $JOB_NAME job runs hourly at :17. Confirm it fired with:"
echo " so-telegraf-repair --check"
+1 -1
View File
@@ -3,7 +3,7 @@
#
# One pillar directory can map to multiple (state, tgt) actions.
# tgt is a raw salt compound expression. tgt_type is always "compound".
# Per-action `batch` / `batch_wait` override the orch defaults (25% / 15s).
# Per-action `batch` / `batch_wait` override the orch defaults (10% / 15s).
# An action with `highstate: True` triggers state.highstate instead of
# state.apply -- see salt/orch/push_batch.sls.
#
+12 -9
View File
@@ -7,7 +7,7 @@
# Two branches, keyed on node_id:
# A) node_id populated -> the change is scoped to that one minion. Look up the
# app in pillar_push_map.yaml and write an intent that runs the app's mapped
# state(s) targeted to just that node.
# state(s) on just that node, and only if the node is in the app's target.
# B) node_id empty -> grid-wide app change. Look up the app in
# pillar_push_map.yaml and write an intent with the entry's actions as-is.
#
@@ -116,15 +116,17 @@ def _app_from_setting(setting_id):
def _node_actions(entry, node_id):
# Copy the app's mapped actions but retarget each one to the single node.
# Copy the app's mapped actions but narrow each one to the single node.
# Preserves the state/highstate selection and any batch/batch_wait overrides.
# The app's own target is kept so the state is not applied to a node that does not
# run the app (e.g. an elasticsearch setting saved for a sensor fails to render there).
actions = []
for action in entry:
if not isinstance(action, dict):
continue
node_action = dict(action)
node_action['tgt'] = node_id
node_action['tgt_type'] = 'glob'
node_action['tgt'] = 'L@{} and ( {} )'.format(node_id, action.get('tgt', '*'))
node_action['tgt_type'] = 'compound'
actions.append(node_action)
return actions
@@ -138,6 +140,7 @@ def run():
# top level so the reactor is robust to either shape.
event = data.get('data', data) # noqa: F821 -- data provided by reactor
setting_id = event.get('setting_id', '')
audit_id = event.get('id')
node_id = (event.get('node_id') or '').strip()
app = _app_from_setting(setting_id)
@@ -150,8 +153,8 @@ def run():
if not entry:
LOG.warning(
'push_pillar: app "%s" is not in pillar_push_map.yaml; change will be '
'picked up at the next scheduled highstate (setting_id=%s)',
app, setting_id,
'picked up at the next scheduled highstate (setting_id=%s audit_id=%s)',
app, setting_id, audit_id,
)
return {}
@@ -165,12 +168,12 @@ def run():
'node_{}_{}'.format(node_id, app), actions,
'audit:{}@{}'.format(setting_id, node_id),
)
LOG.info('push_pillar: per-node intent updated for %s on %s (setting_id=%s)',
app, node_id, setting_id)
LOG.info('push_pillar: per-node intent updated for %s on %s (setting_id=%s audit_id=%s)',
app, node_id, setting_id, audit_id)
return {}
# Branch B: grid-wide app change -> use the map entry's actions as-is.
actions = list(entry) # copy to avoid mutating the cache
_write_intent('pillar_{}'.format(app), actions, 'audit:{}'.format(setting_id))
LOG.info('push_pillar: app intent updated for %s (setting_id=%s)', app, setting_id)
LOG.info('push_pillar: app intent updated for %s (setting_id=%s audit_id=%s)', app, setting_id, audit_id)
return {}
+10
View File
@@ -251,6 +251,16 @@ socsensoronirepos:
- mode: 775
- makedirs: True
nsm_rules_dir:
file.directory:
- name: /nsm/rules
- user: socore
- group: socore
- recurse:
- user
- group
- show_changes: False
create_custom_local_yara_repo_template:
git.present:
- name: /nsm/rules/custom-local-repos/local-yara
+31
View File
@@ -9,6 +9,7 @@
{%- set MDENGINE = GLOBALS.md_engine %}
{%- set LOGSTASH_ENABLED = LOGSTASH_MERGED.enabled %}
{%- set TG_OUT = TELEGRAFMERGED.output | upper %}
{%- set PG_ONLY = TG_OUT == 'POSTGRES' %}
{%- set PG_HOST = GLOBALS.manager_ip %}
{#- Per-minion telegraf creds live in the grid-wide telegraf/creds.sls pillar,
written by /usr/sbin/so-telegraf-cred on the manager. Each minion looks up
@@ -149,7 +150,12 @@
# Read metrics about cpu usage
[[inputs.cpu]]
## Whether to report per-cpu stats or not
{%- if PG_ONLY %}
{#- SOC and so-stats-show read only cpu-total #}
percpu = false
{%- else %}
percpu = true
{%- endif %}
## Whether to report total system cpu stats or not
totalcpu = true
## If true, collect raw CPU time metrics.
@@ -199,6 +205,11 @@
## The typical use case is for LVM volumes, to get the VG/LV name instead of
## the near-meaningless DM-0 name.
# name_templates = ["$ID_FS_LABEL","$DM_VG_NAME/$DM_LV_NAME"]
{%- if PG_ONLY %}
{#- Whole disks, md RAID included, already carry the I/O of their partitions and dm devices #}
[inputs.diskio.tagdrop]
name = ["sd*[0-9]", "vd*[0-9]", "xvd*[0-9]", "hd*[0-9]", "nvme*p[0-9]*", "mmcblk*p[0-9]*", "dm-*", "sr*", "loop*"]
{%- endif %}
# Get kernel statistics from /proc/stat
@@ -242,6 +253,12 @@
password = "{{ ES_PASS }}"
insecure_skip_verify = true
cluster_health = true
{%- if PG_ONLY %}
{#- SOC reads cluster status, indices and clusterstats_nodes ingest stats; alarms plan jvm and fs #}
cluster_health_level = "cluster"
node_stats = ["indices", "jvm", "fs"]
namedrop = ["elasticsearch_clusterstats_indices"]
{%- endif %}
{%- elif grains['role'] in ['so-searchnode'] %}
[[inputs.elasticsearch]]
servers = ["https://{{ NODEIP }}:9200"]
@@ -249,6 +266,10 @@
username = "{{ ES_USER }}"
password = "{{ ES_PASS }}"
insecure_skip_verify = true
{%- if PG_ONLY %}
{#- SOC reads indices; alarms plan jvm and fs #}
node_stats = ["indices", "jvm", "fs"]
{%- endif %}
{%- endif %}
# ## Timeout for HTTP requests to the elastic search server(s)
@@ -291,6 +312,10 @@
collect = ["pipelines"]
username = "{{ salt['pillar.get']('elasticsearch:auth:users:so_logstash_user:user') }}"
password = "{{ salt['pillar.get']('elasticsearch:auth:users:so_logstash_user:pass') }}"
{%- if PG_ONLY %}
{#- SOC reads logstash_events only #}
namedrop = ["logstash_plugins"]
{%- endif %}
{%- endif %}
{% if grains.role in ['so-manager','so-managerhype','so-managersearch','so-standalone','so-receiver'] and GLOBALS.pipeline == "KAFKA" -%}
@@ -392,6 +417,12 @@
# # Read metrics about network interface usage
[[inputs.net]]
{%- if PG_ONLY %}
{%- from 'docker/docker.map.jinja' import DOCKERMERGED %}
{#- Only virtual interfaces; every physical NIC, bond members included, is kept for SOC and alarms #}
[inputs.net.tagdrop]
interface = {{ (['veth*', 'docker*', 'lo'] + DOCKERMERGED.networks.keys() | list) | json }}
{%- endif %}
# Scripts run every 30s||TELEGRAFMERGED.config.interval - ES index script doesn't need to run as frequently
{%- if GLOBALS.is_manager or GLOBALS.role == 'so-heavynode' %}