Telegraf: fix EPS double counting and duplicate metric rows

surface ilm policy errors in so-elasticsearch-ilm-policy-load output

.github/DISCUSSION_TEMPLATE/3-0.yml

@@ -14,6 +14,7 @@ body:
         - 3.2.0
         - 3.3.0
         - 3.4.0
+        - 3.5.0
         - Other (please provide detail below)
     validations:
       required: true

VERSION

@@ -1 +1 @@
-3.4.0
+3.5.0

salt/desktop/packages.sls

@@ -438,6 +438,10 @@ install_networkminer:
   pkg.latest:
     - name: securityonion-networkminer
 
+install_so_crates:
+  pkg.latest:
+    - name: securityonion-so-crates
+
 {% else %}
 
 desktop_packages_os_fail:

salt/elasticsearch/cluster.sls

@@ -107,6 +107,12 @@ so-elasticsearch-system-indices-patch:
       - file: so-elasticsearch-system-indices-patch-script
 {%  endif %}
 
+so-elasticsearch-ilm-policy-reload:
+  file.absent:
+    - name: /opt/so/state/esilm.txt
+    - onchanges:
+      - file: so-elasticsearch-ilm-policy-load-script
+
 # heavynodes will only load ILM policies for SO managed indices. (Indicies defined in elasticsearch/defaults.yaml)
 so-elasticsearch-ilm-policy-load:
   cmd.run:
@@ -115,8 +121,9 @@ so-elasticsearch-ilm-policy-load:
     - require:
       - docker_container: so-elasticsearch
       - file: so-elasticsearch-ilm-policy-load-script
-    - onchanges:
-      - file: so-elasticsearch-ilm-policy-load-script
+      - http: wait_for_so-elasticsearch
+      - file: so-elasticsearch-ilm-policy-reload
+    - unless: test -f /opt/so/state/esilm.txt
 
 so-elasticsearch-templates-reload:
   file.absent:

salt/elasticsearch/tools/sbin_jinja/so-elasticsearch-ilm-policy-load

@@ -7,11 +7,17 @@
 . /usr/sbin/so-common
 
 MAX_JOBS=${MAX_ILM_JOBS:-10}
+ILM_STATEFILE=/opt/so/state/esilm.txt
+rm -f "$ILM_STATEFILE"
 
+check_elasticsearch_responsive
+
+ILM_TMP_DIR=$(mktemp -d)
+trap 'rm -rf "$ILM_TMP_DIR"' EXIT
 # Lock used to serialize block writes so concurrent jobs never interleave their output.
-ILM_OUTPUT_LOCK=$(mktemp)
-ILM_FAIL_FILE=$(mktemp)
-trap 'rm -f "$ILM_OUTPUT_LOCK" "$ILM_FAIL_FILE"' EXIT
+ILM_OUTPUT_LOCK="$ILM_TMP_DIR/output.lock"
+ILM_FAIL_FILE="$ILM_TMP_DIR/failed"
+touch "$ILM_FAIL_FILE"
 
 # Policies are loaded concurrently (up to MAX_JOBS at a time) for speed. Each policy's block is
 # printed the moment its curl returns, so output appears in COMPLETION ORDER, not the order
@@ -20,13 +26,30 @@ echo "Loading ILM policies concurrently; output below appears in completion orde
 echo
 
 put_policy() {
-  local desc="$1" policyname="$2" data="$3" result rc=0
-  if ! result=$(curl -K /opt/so/conf/elasticsearch/curl.config -s -k -L --fail \
-    -X PUT "https://localhost:9200/_ilm/policy/${policyname}" \
-    -H 'Content-Type: application/json' -d"${data}" 2>&1); then
-    rc=1
-  elif ! jq -e '.acknowledged == true' <<<"$result" >/dev/null 2>&1; then
-    rc=1
+  local desc="$1" policyname="$2" data="$3" rc=1 attempt out http_code body result reason tmpdata
+  tmpdata=$(mktemp "$ILM_TMP_DIR/${policyname}.XXXXXX.json")
+  printf '%s' "$data" >"$tmpdata"
+
+  for attempt in 1 2 3; do
+    out=$(so-elasticsearch-query "_ilm/policy/${policyname}" -d@"${tmpdata}" -XPUT -w '\n%{http_code}' 2>/dev/null)
+    http_code=${out##*$'\n'} body=${out%$'\n'*}
+
+    if [[ "$http_code" == 200 ]] && jq -e '.acknowledged == true' <<<"$body" >/dev/null 2>&1; then
+      rc=0
+      break
+    fi
+    # 429 = ES rejected/throttled (may succeed later); other 4xx will not
+    [[ "$http_code" == 4?? && "$http_code" != 429 ]] && break
+    (( attempt < 3 )) && sleep 3
+  done
+
+  if (( rc == 0 )); then
+    result='{"acknowledged":true}'
+  else
+    result=${body:-"HTTP ${http_code:-000}"}
+    # flatten tabs/newlines so the TAB fail-file stays one record
+    reason=$(jq -r '(if (.error|type) == "object" then (.error.reason // .error.type) elif (.error|type) == "string" then .error else empty end) // empty | tostring | gsub("[\t\r\n]+"; " ")' <<<"$body" 2>/dev/null)
+    reason="HTTP ${http_code:-000}${reason:+: ${reason}}"
   fi
 
   # curl above ran in parallel; serialize just this block write so concurrent jobs never interleave.
@@ -34,7 +57,7 @@ put_policy() {
     flock 200
     printf 'Setting up %s policy...\n%s\n\n' "${desc}" "${result}"
     if (( rc != 0 )); then
-      printf '%s\n' "${policyname}" >>"$ILM_FAIL_FILE"
+      printf '%s\t%s\n' "${policyname}" "${reason}" >>"$ILM_FAIL_FILE"
     fi
   } 200>>"${ILM_OUTPUT_LOCK}"
 
@@ -82,10 +105,15 @@ wait || true
 
 if [[ -s "$ILM_FAIL_FILE" ]]; then
   echo "ERROR: Failed to load ILM policy(s):"
-  while read -r POLICY; do
-    echo " - $POLICY"
+  while IFS=$'\t' read -r POLICY REASON; do
+    if [[ -n "$REASON" ]]; then
+      echo " - ${POLICY}: ${REASON}"
+    else
+      echo " - ${POLICY}"
+    fi
   done < "$ILM_FAIL_FILE"
   exit 1
 else
   echo "Successfully loaded all ILM policies."
+  touch "$ILM_STATEFILE"
 fi

salt/manager/tools/sbin/soup

@@ -121,9 +121,6 @@ check_err() {
       161)
         echo 'Required intermediate Elasticsearch upgrade not complete'
       ;;
-      162)
-        echo 'One or more Elastic Agent nodes do not support the x86-64-v3 CPU instruction set'
-      ;;
       170)
         echo "Intermediate upgrade completed successfully to $next_step_so_version, but next soup to Security Onion $originally_requested_so_version could not be started automatically."
         echo "Start soup again manually to continue the upgrade to Security Onion $originally_requested_so_version."
@@ -351,83 +348,6 @@ check_cluster_health() {
   exit 0
 }
 
-no_soup_for_you() {
-  echo ""
-  echo "No soup for you!"
-  exit 162
-}
-
-check_cpu_compatibility() {
-  # Roles running a container built from the so-elastic-agent image; mirrors the
-  # elasticagent and elasticfleet entries in salt/reactor/pillar_push_map.yaml.
-  local cpu_target='G@role:so-heavynode or G@role:so-eval or G@role:so-fleet or G@role:so-import or G@role:so-manager or G@role:so-managerhype or G@role:so-managersearch or G@role:so-standalone'
-  local expected_nodes cpu_results node result confirm
-  local -a unsupported=() offline=()
-
-  echo "Checking that Elastic Agent nodes support the x86-64-v3 CPU instruction set now required by Elastic."
-
-  if [[ "$SKIP_CPU_CHECK" == "true" ]]; then
-    printf "\nSkipping the x86-64-v3 CPU check because --skip-cpu-check was specified.\n\n"
-    return
-  fi
-
-  # Nodes that never answer are absent from the results, so diff against who should have.
-  expected_nodes=$(salt -C "$cpu_target" --preview-target --out=json 2>/dev/null | jq -r '.[]?') || true
-  if [[ -z "$expected_nodes" ]]; then
-    printf "\nCould not determine which nodes run the Elastic Agent, so the x86-64-v3 CPU check cannot run.\n"
-    no_soup_for_you
-  fi
-
-  cpu_results=$(salt -t 30 -C "$cpu_target" cmd.run "/lib64/ld-linux-x86-64.so.2 --help | grep x86-64-v3" --out=json 2>/dev/null) || true
-
-  while IFS= read -r node; do
-    [[ -z "$node" ]] && continue
-    result=$(jq -r --arg node "$node" '.[$node] // empty' <<< "$cpu_results" 2>/dev/null)
-    if [[ -z "$result" || "$result" == *"did not return"* ]]; then
-      offline+=("$node")
-    elif [[ "$result" != *"x86-64-v3 (supported"* ]]; then
-      # glibc appends "(supported, searched)" only when supported; the open paren keeps
-      # this from matching a future "(unsupported".
-      unsupported+=("$node")
-    fi
-  done <<< "$expected_nodes"
-
-  if [[ ${#unsupported[@]} -eq 0 && ${#offline[@]} -eq 0 ]]; then
-    printf "\nAll Elastic Agent nodes support x86-64-v3. We can proceed with SOUP.\n\n"
-    return
-  fi
-
-  echo ""
-  if [[ ${#unsupported[@]} -gt 0 ]]; then
-    echo "The following node(s) do NOT support the x86-64-v3 CPU instruction set:"
-    printf '  %s\n' "${unsupported[@]}"
-    echo ""
-    echo "Upstream Elastic now builds its binaries for x86-64-v3, so these nodes can no"
-    echo "longer run Elastic. Upgrading them WILL BREAK them."
-    echo ""
-  fi
-  if [[ ${#offline[@]} -gt 0 ]]; then
-    echo "The following node(s) did not respond and could not be checked:"
-    printf '  %s\n' "${offline[@]}"
-    echo ""
-    echo "These nodes are offline, so we cannot confirm they support x86-64-v3, which"
-    echo "upstream Elastic now requires."
-    echo ""
-  fi
-
-  if [[ -n $UNATTENDED ]]; then
-    echo "Unattended mode cannot prompt for an override. Re-run soup interactively, or pass --skip-cpu-check to bypass this check."
-    no_soup_for_you
-  fi
-
-  read -rp "Type 'override' to continue anyway, or press Enter to exit: " confirm
-  if [[ "${confirm,,}" == "override" ]]; then
-    printf "\nOverride accepted. Continuing at your own risk.\n\n"
-  else
-    no_soup_for_you
-  fi
-}
-
 check_fleet_server() {
   echo "Checking that Elastic Fleet Server is responding."
   # Modeled on the wait_for_so-elastic-fleet state check in elasticfleet/enabled.sls,
@@ -2168,8 +2088,6 @@ main() {
   echo "Let's see if we need to update Security Onion."
   upgrade_check
 
-  check_cpu_compatibility
-
   upgrade_space
 
   echo "Verifying Elasticsearch version compatibility across the grid before upgrading."
@@ -2448,17 +2366,6 @@ fi
   echo "### soup has been served at $(date) ###"
 }
 
-SKIP_CPU_CHECK=false
-declare -a SOUP_ARGS=()
-for arg in "$@"; do
-  if [[ "$arg" == "--skip-cpu-check" ]]; then
-    SKIP_CPU_CHECK=true
-  else
-    SOUP_ARGS+=("$arg")
-  fi
-done
-set -- "${SOUP_ARGS[@]}"
-
 while getopts ":b:f:y" opt; do
   case ${opt} in
     b )
@@ -2482,7 +2389,7 @@ while getopts ":b:f:y" opt; do
       ISOLOC="$OPTARG"
     ;;
     \? )
-      echo "Usage: soup [-b] [-y] [-f <iso location>] [--skip-cpu-check]"
+      echo "Usage: soup [-b] [-y] [-f <iso location>]"
       exit 1
     ;;
     : )

salt/telegraf/etc/telegraf.conf

@@ -138,6 +138,10 @@
 #                            AGGREGATOR PLUGINS                               #
 ###############################################################################
 
+[[aggregators.merge]]
+  drop_original = true
+  namepass = ["system", "swap", "kafka_partition"]
+
 ###############################################################################
 #                            INPUT PLUGINS                                    #
 ###############################################################################
@@ -289,7 +293,7 @@
   password = "{{ salt['pillar.get']('elasticsearch:auth:users:so_logstash_user:pass') }}"
 {%- endif %}
 
-{% if grains.role in ['so-manager','so-managersearch','so-standalone','so-receiver'] and GLOBALS.pipeline == "KAFKA" -%}
+{% if grains.role in ['so-manager','so-managerhype','so-managersearch','so-standalone','so-receiver'] and GLOBALS.pipeline == "KAFKA" -%}
 [[inputs.jolokia2_agent]]
   name_prefix= "kafka_"
   urls = ["http://{{ NODEIP }}:8778/jolokia"]
@@ -346,6 +350,26 @@
    interval = "60s"
 {%- endif %}
 
+{#- These report state that changes slowly; collecting them every interval only stores duplicate rows #}
+{%- set SLOWSCRIPTS = [] %}
+{%- for script in ['agentstatus.sh', 'os.sh', 'features.sh', 'raid.sh', 'zeekloss.sh'] %}
+{%-   if script in TELEGRAFMERGED.scripts[GLOBALS.role.split('-')[1]] %}
+{%-     do TELEGRAFMERGED.scripts[GLOBALS.role.split('-')[1]].remove(script) %}
+{%-     do SLOWSCRIPTS.append(script) %}
+{%-   endif %}
+{%- endfor %}
+{%- if SLOWSCRIPTS | length > 0 %}
+[[inputs.exec]]
+   commands = [
+{%-   for script in SLOWSCRIPTS %}
+      ["/scripts/{{script}}"]{% if not loop.last %},{% endif %}
+{%-   endfor %}
+   ]
+   data_format = "influx"
+   timeout = "15s"
+   interval = "5m"
+{%- endif %}
+
 {%- if TELEGRAFMERGED.scripts[GLOBALS.role.split('-')[1]] | length > 0 %}
 [[inputs.exec]]
    commands = [
@@ -358,12 +382,6 @@
   timeout = "15s"
 {%- endif %}
 
-{%- if salt['pillar.get']('healthcheck:enabled', False) %}
-[[inputs.file]]
-  files = ["/host/nsm/zeek/logs/zeek_restart.log"]
-  data_format = "influx"
-{%- endif %}
-
 [[inputs.file]]
   files = ["/etc/telegraf/node_config.json"]
   name_override = "node_config"

salt/telegraf/map.jinja

@@ -18,7 +18,7 @@
 
 {% if GLOBALS.pipeline != 'REDIS' %}
 {# When global pipeline is not REDIS remove redis.sh script. KAFKA metrics are collected via jolokia agent. Config in telegraf.conf #}
-{%   if GLOBALS.role in ['so-standalone', 'so-manager', 'so-managersearch', 'so-receiver', 'so-heavynode'] %}
+{%   if GLOBALS.role in ['so-standalone', 'so-manager', 'so-managerhype', 'so-managersearch', 'so-receiver', 'so-heavynode'] %}
 {%     do TELEGRAFMERGED.scripts[GLOBALS.role.split('-')[1]].remove('redis.sh') %}
 {%   endif %}
 {% endif %}

salt/telegraf/scripts/checkfiles.sh

@@ -10,7 +10,7 @@
 # if this script isn't already running
 if [[ ! "`pidof -x $(basename $0) -o %PPID`" ]]; then
 
-    FILES=$(ls -1x /host/nsm/strelka/unprocessed | wc -l)
+    FILES=$(find /host/nsm/strelka/unprocessed -maxdepth 1 -type f | wc -l)
     echo "faffiles files=$FILES"
     
 fi

salt/telegraf/scripts/containers.sh

@@ -12,8 +12,9 @@ if [[ ! "`pidof -x $(basename $0) -o %PPID`" ]]; then
 
     CONTAINERSLOG=/var/log/somon/containers.log
     # the collector rewrites this every minute; report nothing rather than repeating a stale
-    # file as if it were current, in case a run was skipped or the collector is wedged
-    MAXAGE=150
+    # file as if it were current, in case a run was skipped or the collector is wedged.
+    # Kept under two collector cycles so a 60s telegraf interval can't emit the same snapshot twice.
+    MAXAGE=90
 
     if [ -r "$CONTAINERSLOG" ]; then
         AGE=$(( $(date +%s) - $(stat -c %Y "$CONTAINERSLOG") ))

salt/telegraf/scripts/eps.sh

@@ -11,27 +11,40 @@
 if [[ ! "`pidof -x $(basename $0) -o %PPID`" ]]; then
 
     PREVCOUNTFILE='/tmp/eps.txt'
-    EVENTCOUNTCURRENT="$(curl -s localhost:9600/_node/stats | jq '.events.in')"
+    STATS="$(curl -s localhost:9600/_node/stats)"
 
-    if [ ! -z "$EVENTCOUNTCURRENT" ]; then
-
-      if [ -f "$PREVCOUNTFILE" ]; then
-        EVENTCOUNTPREVIOUS=`cat $PREVCOUNTFILE`
-      else
-        echo "${EVENTCOUNTCURRENT}" > $PREVCOUNTFILE
-        exit 0
-      fi
-
-      echo "${EVENTCOUNTCURRENT}" > $PREVCOUNTFILE
-      # the division by 30 is because the agent interval is 30 seconds
-      EVENTS=$(((EVENTCOUNTCURRENT - EVENTCOUNTPREVIOUS)/30))
-      if [ "$EVENTS" -lt 0 ]; then
-        EVENTS=0
-      fi
-
-      echo "consumptioneps eps=${EVENTS%%.*}"
+    if echo "$STATS" | jq -e '.pipelines.search' > /dev/null 2>&1; then
+      STAGE=search
+      COUNT="$(echo "$STATS" | jq '.pipelines.search.events.out')"
+    elif echo "$STATS" | jq -e '.pipelines.receiver' > /dev/null 2>&1; then
+      STAGE=receiver
+      COUNT="$(echo "$STATS" | jq '.pipelines.receiver.events.in')"
+    else
+      exit 0
     fi
 
+    if ! [[ "$COUNT" =~ ^[0-9]+$ ]]; then
+      exit 0
+    fi
+
+    NOW=$(date +%s)
+    read -r PREVEPOCH PREVCOUNT 2>/dev/null < "$PREVCOUNTFILE"
+    echo "${NOW} ${COUNT}" > "$PREVCOUNTFILE"
+
+    if ! [[ "$PREVEPOCH" =~ ^[0-9]+$ && "$PREVCOUNT" =~ ^[0-9]+$ ]]; then
+      exit 0
+    fi
+
+    ELAPSED=$((NOW - PREVEPOCH))
+    DELTA=$((COUNT - PREVCOUNT))
+    if [ "$ELAPSED" -le 0 ] || [ "$DELTA" -lt 0 ]; then
+      EPS=0
+    else
+      EPS=$((DELTA / ELAPSED))
+    fi
+
+    echo "consumptioneps,stage=${STAGE} eps=${EPS}"
+
 fi
 
 exit 0

salt/telegraf/scripts/features.sh

@@ -10,8 +10,7 @@ if [[ ! "`pidof -x $(basename $0) -o %PPID`" ]]; then
     FPS_ENABLED=$(cat /var/log/sostatus/fps_enabled)
     LKS_ENABLED=$(cat /var/log/sostatus/lks_enabled)
 
-    echo "features fps=$FPS_ENABLED"
-    echo "features lks=$LKS_ENABLED"
+    echo "features fps=$FPS_ENABLED,lks=$LKS_ENABLED"
 fi
 
 exit 0

setup/so-functions

@@ -976,27 +976,50 @@ docker_seed_update() {
 	set_progress_str "$docker_seed_update_percent" "Downloading $name"
 }
 
+registry_tar_has_soversion() {
+	local tarfile=$1
+	local version=$2
+
+	[ -n "$version" ] || return 1
+	[ -f "$tarfile" ] || return 1
+	# grep -q in a pipe SIGPIPEs tar; under pipefail a match would look like failure.
+	grep -qF "/_manifests/tags/${version}/" < <(tar -tf "$tarfile" 2>/dev/null)
+}
+
 docker_seed_registry() {
 	local VERSION="$SOVERSION"
+	local REGISTRY_TAR=/nsm/docker-registry/docker/registry.tar
 
-	if [ -f /nsm/docker-registry/docker/registry.tar ]; then
-		logCmd "tar xvf /nsm/docker-registry/docker/registry.tar -C /nsm/docker-registry/docker"
-		logCmd "rm /nsm/docker-registry/docker/registry.tar"
+	if [ -f "$REGISTRY_TAR" ] && registry_tar_has_soversion "$REGISTRY_TAR" "$VERSION"; then
+		info "Extracting $REGISTRY_TAR"
+		if ! tar xvf "$REGISTRY_TAR" -C /nsm/docker-registry/docker >> "$setup_log" 2>&1; then
+			error "Failed to extract $REGISTRY_TAR"
+			return 1
+		fi
+		logCmd "rm -f $REGISTRY_TAR"
+		return 0
+	elif [ -f "$REGISTRY_TAR" ]; then
+		info "Not using $REGISTRY_TAR; it has no images tagged '${VERSION}'"
+		if [[ "$is_airgap" == true ]]; then
+			error "Airgap install: $REGISTRY_TAR does not match Security Onion version '${VERSION}'. Use the ISO for this version."
+			return 1
+		fi
 	elif [[ -d /nsm/docker-registry/docker/registry && ( -f /etc/SOCLOUD || "$is_airgap" == true ) ]]; then
 		echo "Using existing docker registry content"
-	else
-		if [ "$install_type" == 'IMPORT' ]; then
-			container_list 'so-import'	
-		else
-			container_list
-		fi
-
-		docker_seed_update_percent=25
-
-		update_docker_containers 'netinstall' '' 'docker_seed_update' '/dev/stdout' 2>&1 | tee -a "$setup_log"
-        # Use pipe exit status of 'update_docker_containers' for return code
-		return ${PIPESTATUS[0]}
+		return 0
 	fi
+
+	if [ "$install_type" == 'IMPORT' ]; then
+		container_list 'so-import'	
+	else
+		container_list
+	fi
+
+	docker_seed_update_percent=25
+
+	update_docker_containers 'netinstall' '' 'docker_seed_update' '/dev/stdout' 2>&1 | tee -a "$setup_log"
+        # Use pipe exit status of 'update_docker_containers' for return code
+	return ${PIPESTATUS[0]}
 }
 
 elasticfleet_pillar() {