Comingle Telegraf metrics into shared schema

Per-minion schemas cause table count to explode (N minions * M metrics)
and the per-minion revocation story isn't worth it when retention is
short. Move all minions to a shared 'telegraf' schema while keeping
per-minion login credentials for audit.

- New so_telegraf NOLOGIN group role owns the telegraf schema; each
  per-minion role is a member and inherits insert/select via role
  inheritance
- Telegraf connection string uses options='-c role=so_telegraf' so
  tables auto-created on first write belong to the group role
- so-telegraf-trim walks the flat telegraf.* table set instead of
  per-minion schemas
- so-stats-show filters by host tag; CLI arg is now the hostname as
  tagged by Telegraf rather than a sanitized schema suffix
- Also renames so-show-stats -> so-stats-show
This commit is contained in:
Mike Reeves
2026-04-16 15:40:54 -04:00
parent c124186989
commit 470b3bd4da
5 changed files with 204 additions and 154 deletions
+24 -2
View File
@@ -10,6 +10,28 @@
{% set TG_OUT = (GLOBALS.telegraf_output | default('INFLUXDB')) | upper %}
{% if TG_OUT in ['POSTGRES', 'BOTH'] %}
# Provision the shared group role and schema once. Every per-minion role is a
# member of so_telegraf, and each Telegraf connection does SET ROLE so_telegraf
# (via options='-c role=so_telegraf' in the connection string) so tables created
# on first write are owned by the group role and every member can INSERT/SELECT.
postgres_telegraf_group_role:
cmd.run:
- name: |
docker exec -i so-postgres psql -v ON_ERROR_STOP=1 -U postgres -d so_telegraf <<'EOSQL'
DO $$
BEGIN
IF NOT EXISTS (SELECT FROM pg_catalog.pg_roles WHERE rolname = 'so_telegraf') THEN
CREATE ROLE so_telegraf NOLOGIN;
END IF;
END
$$;
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;
EOSQL
- require:
- docker_container: so-postgres
{% set users = salt['pillar.get']('postgres:auth:users', {}) %}
{% for key, entry in users.items() %}
{% if key.startswith('telegraf_') and entry.get('user') and entry.get('pass') %}
@@ -30,10 +52,10 @@ postgres_telegraf_role_{{ u }}:
END
$$;
GRANT CONNECT ON DATABASE so_telegraf TO "{{ u }}";
CREATE SCHEMA IF NOT EXISTS "{{ u }}" AUTHORIZATION "{{ u }}";
GRANT so_telegraf TO "{{ u }}";
EOSQL
- require:
- docker_container: so-postgres
- cmd: postgres_telegraf_group_role
{% endif %}
{% endfor %}
-145
View File
@@ -1,145 +0,0 @@
#!/bin/bash
# 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.
# Point-in-time host metrics from the Telegraf Postgres backend.
# Sanity-check tool for verifying metrics are landing before the grid
# dashboards consume them.
. /usr/sbin/so-common
usage() {
cat <<EOF
Usage: $0 [minion_id]
Shows the most recent CPU, memory, disk, and load metrics for each minion
from the so_telegraf Postgres database. Without an argument, reports on
every minion that has data. With a minion_id, limits output to that one.
Requires: sudo, so-postgres running, global.telegraf_output set to
POSTGRES or BOTH.
EOF
exit 1
}
if [ "$(id -u)" -ne 0 ]; then
echo "This script must be run using sudo!"
exit 1
fi
case "${1:-}" in
-h|--help) usage ;;
esac
FILTER_MINION="${1:-}"
so_psql() {
docker exec so-postgres psql -U postgres -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
echo "Database so_telegraf not found. Is global.telegraf_output set to POSTGRES or BOTH?"
exit 2
fi
# List telegraf schemas (role-per-minion naming convention: so_telegraf_<sanitized_minion_id>)
SCHEMAS=$(so_psql -c "SELECT schema_name FROM information_schema.schemata WHERE schema_name LIKE 'so_telegraf_%' ORDER BY schema_name;")
if [ -z "$SCHEMAS" ]; then
echo "No minion schemas found in so_telegraf."
exit 0
fi
print_metric() {
local schema="$1" table="$2" query="$3"
# Confirm table exists in this schema before querying
local exists
exists=$(so_psql -c "SELECT 1 FROM information_schema.tables WHERE table_schema='${schema}' AND table_name='${table}' LIMIT 1;")
[ -z "$exists" ] && return 0
so_psql -c "$query"
}
# Telegraf's postgresql output stores tag values either as individual columns
# on the <metric>_tag table or as a single JSONB "tags" column, depending on
# plugin version. Returns a SQL expression that extracts the named tag
# regardless of layout. Empty string if the tag table doesn't exist.
tag_expr() {
local schema="$1" table="$2" tag="$3" alias="$4"
local has_col
has_col=$(so_psql -c "
SELECT 1 FROM information_schema.columns
WHERE table_schema='${schema}' AND table_name='${table}_tag' AND column_name='${tag}'
LIMIT 1;")
if [ -n "$has_col" ]; then
echo "${alias}.${tag}"
return
fi
local has_tags
has_tags=$(so_psql -c "
SELECT 1 FROM information_schema.columns
WHERE table_schema='${schema}' AND table_name='${table}_tag' AND column_name='tags'
LIMIT 1;")
if [ -n "$has_tags" ]; then
echo "(${alias}.tags->>'${tag}')"
return
fi
echo ""
}
for schema in $SCHEMAS; do
minion="${schema#so_telegraf_}"
if [ -n "$FILTER_MINION" ]; then
# Compare against the sanitized form used in schema names
want=$(echo "$FILTER_MINION" | tr '.-' '_' | tr '[:upper:]' '[:lower:]')
[ "$minion" != "$want" ] && continue
fi
echo "===================================================================="
echo " Minion: $minion"
echo "===================================================================="
cpu_tag=$(tag_expr "$schema" "cpu" "cpu" "t")
if [ -n "$cpu_tag" ]; then
print_metric "$schema" "cpu" "
SELECT 'cpu ' AS metric,
to_char(c.time, 'YYYY-MM-DD HH24:MI:SS') AS ts,
round((100 - c.usage_idle)::numeric, 1) || '% used'
FROM \"${schema}\".cpu c
JOIN \"${schema}\".cpu_tag t USING (tag_id)
WHERE ${cpu_tag} = 'cpu-total'
ORDER BY c.time DESC LIMIT 1;"
fi
print_metric "$schema" "mem" "
SELECT 'memory ' AS metric,
to_char(m.time, 'YYYY-MM-DD HH24:MI:SS') AS ts,
round(m.used_percent::numeric, 1) || '% used (' ||
pg_size_pretty(m.used) || ' of ' || pg_size_pretty(m.total) || ')'
FROM \"${schema}\".mem m
ORDER BY m.time DESC LIMIT 1;"
disk_path=$(tag_expr "$schema" "disk" "path" "t")
if [ -n "$disk_path" ]; then
print_metric "$schema" "disk" "
SELECT 'disk ' || rpad(${disk_path}, 12) AS metric,
to_char(d.time, 'YYYY-MM-DD HH24:MI:SS') AS ts,
round(d.used_percent::numeric, 1) || '% used (' ||
pg_size_pretty(d.used) || ' of ' || pg_size_pretty(d.total) || ')'
FROM \"${schema}\".disk d
JOIN \"${schema}\".disk_tag t USING (tag_id)
WHERE d.time = (SELECT max(time) FROM \"${schema}\".disk)
ORDER BY ${disk_path};"
fi
print_metric "$schema" "system" "
SELECT 'load ' AS metric,
to_char(s.time, 'YYYY-MM-DD HH24:MI:SS') AS ts,
s.load1 || ' / ' || s.load5 || ' / ' || s.load15 || ' (1/5/15m)'
FROM \"${schema}\".system s
ORDER BY s.time DESC LIMIT 1;"
echo ""
done
+170
View File
@@ -0,0 +1,170 @@
#!/bin/bash
# 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.
# Point-in-time host metrics from the Telegraf Postgres backend.
# Sanity-check tool for verifying metrics are landing before the grid
# dashboards consume them.
. /usr/sbin/so-common
usage() {
cat <<EOF
Usage: $0 [minion_id]
Shows the most recent CPU, memory, disk, and load metrics for each minion
from the so_telegraf Postgres database. Without an argument, reports on
every minion that has data. With a minion_id, limits output to that one.
Requires: sudo, so-postgres running, global.telegraf_output set to
POSTGRES or BOTH.
EOF
exit 1
}
if [ "$(id -u)" -ne 0 ]; then
echo "This script must be run using sudo!"
exit 1
fi
case "${1:-}" in
-h|--help) usage ;;
esac
FILTER_HOST="${1:-}"
SCHEMA="telegraf"
so_psql() {
docker exec so-postgres psql -U postgres -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
echo "Database so_telegraf not found. Is global.telegraf_output set to POSTGRES or BOTH?"
exit 2
fi
# Telegraf's postgresql output stores tag values either as individual columns
# on the <metric>_tag table or as a single JSONB "tags" column, depending on
# plugin version. Returns a SQL expression that extracts the named tag
# regardless of layout. Empty string if the tag table doesn't exist.
tag_expr() {
local table="$1" tag="$2" alias="$3"
local has_col
has_col=$(so_psql -c "
SELECT 1 FROM information_schema.columns
WHERE table_schema='${SCHEMA}' AND table_name='${table}_tag' AND column_name='${tag}'
LIMIT 1;")
if [ -n "$has_col" ]; then
echo "${alias}.${tag}"
return
fi
local has_tags
has_tags=$(so_psql -c "
SELECT 1 FROM information_schema.columns
WHERE table_schema='${SCHEMA}' AND table_name='${table}_tag' AND column_name='tags'
LIMIT 1;")
if [ -n "$has_tags" ]; then
echo "(${alias}.tags->>'${tag}')"
return
fi
echo ""
}
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;")" ]
}
# Discover hosts from cpu_tag (every minion reports cpu).
host_expr=$(tag_expr "cpu" "host" "t")
if [ -z "$host_expr" ]; then
echo "Unable to determine host tag column on ${SCHEMA}.cpu_tag. Has Telegraf written any rows yet?"
exit 0
fi
HOSTS=$(so_psql -c "
SELECT DISTINCT ${host_expr}
FROM \"${SCHEMA}\".cpu_tag t
WHERE ${host_expr} IS NOT NULL
ORDER BY 1;")
if [ -z "$HOSTS" ]; then
echo "No hosts found in ${SCHEMA}. Is Telegraf configured to write to Postgres?"
exit 0
fi
print_metric() {
local query="$1"
so_psql -c "$query"
}
for host in $HOSTS; do
if [ -n "$FILTER_HOST" ] && [ "$host" != "$FILTER_HOST" ]; then
continue
fi
echo "===================================================================="
echo " Host: $host"
echo "===================================================================="
cpu_host=$(tag_expr "cpu" "host" "t")
cpu_tag=$(tag_expr "cpu" "cpu" "t")
if [ -n "$cpu_host" ] && [ -n "$cpu_tag" ]; then
print_metric "
SELECT 'cpu ' AS metric,
to_char(c.time, 'YYYY-MM-DD HH24:MI:SS') AS ts,
round((100 - c.usage_idle)::numeric, 1) || '% used'
FROM \"${SCHEMA}\".cpu c
JOIN \"${SCHEMA}\".cpu_tag t USING (tag_id)
WHERE ${cpu_host} = '${host}' AND ${cpu_tag} = 'cpu-total'
ORDER BY c.time DESC LIMIT 1;"
fi
mem_host=$(tag_expr "mem" "host" "t")
if [ -n "$mem_host" ] && table_exists "mem"; then
print_metric "
SELECT 'memory ' AS metric,
to_char(m.time, 'YYYY-MM-DD HH24:MI:SS') AS ts,
round(m.used_percent::numeric, 1) || '% used (' ||
pg_size_pretty(m.used) || ' of ' || pg_size_pretty(m.total) || ')'
FROM \"${SCHEMA}\".mem m
JOIN \"${SCHEMA}\".mem_tag t USING (tag_id)
WHERE ${mem_host} = '${host}'
ORDER BY m.time DESC LIMIT 1;"
fi
disk_host=$(tag_expr "disk" "host" "t")
disk_path=$(tag_expr "disk" "path" "t")
if [ -n "$disk_host" ] && [ -n "$disk_path" ] && table_exists "disk"; then
print_metric "
SELECT 'disk ' || rpad(${disk_path}, 12) AS metric,
to_char(d.time, 'YYYY-MM-DD HH24:MI:SS') AS ts,
round(d.used_percent::numeric, 1) || '% used (' ||
pg_size_pretty(d.used) || ' of ' || pg_size_pretty(d.total) || ')'
FROM \"${SCHEMA}\".disk d
JOIN \"${SCHEMA}\".disk_tag t USING (tag_id)
WHERE ${disk_host} = '${host}'
AND d.time = (SELECT max(d2.time)
FROM \"${SCHEMA}\".disk d2
JOIN \"${SCHEMA}\".disk_tag t2 USING (tag_id)
WHERE ${disk_host/t./t2.} = '${host}')
ORDER BY ${disk_path};"
fi
sys_host=$(tag_expr "system" "host" "t")
if [ -n "$sys_host" ] && table_exists "system"; then
print_metric "
SELECT 'load ' AS metric,
to_char(s.time, 'YYYY-MM-DD HH24:MI:SS') AS ts,
s.load1 || ' / ' || s.load5 || ' / ' || s.load15 || ' (1/5/15m)'
FROM \"${SCHEMA}\".system s
JOIN \"${SCHEMA}\".system_tag t USING (tag_id)
WHERE ${sys_host} = '${host}'
ORDER BY s.time DESC LIMIT 1;"
fi
echo ""
done
+4 -4
View File
@@ -63,15 +63,15 @@ log "Trimming rows older than ${DAYS} days (dry_run=${DRY_RUN})."
TOTAL_DELETED=0
# One row per (schema, table) we might want to trim.
# Column name is 'time' for all telegraf output plugin tables; skip metadata
# tables (tag_* used for tags_as_foreign_keys).
# Every metric table in the shared telegraf schema has a 'time' column.
# Tag tables (<metric>_tag) don't, so filtering on the column presence is
# enough to scope the trim to metric tables only.
ROWS=$(so_psql -c "
SELECT table_schema || '.' || table_name
FROM information_schema.columns
WHERE column_name = 'time'
AND data_type IN ('timestamp with time zone', 'timestamp without time zone')
AND table_schema LIKE 'so_telegraf_%'
AND table_schema = 'telegraf'
ORDER BY 1;")
if [ -z "$ROWS" ]; then