:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.
Python ab-testingai-analyticsanalyticscdpdata-warehouseexperimentsfeature-flagsjavascriptproduct-analyticspythonreactsession-replaysurveystypescriptweb-analytics
#!/bin/sh
# Renders the OpenTelemetry Collector config for the PostHog metrics agent,
# then execs the collector. See README.md for the env var surface.
#
# Config resolution order. Sources are additive: prometheus scraping and
# Google Cloud Monitoring can both feed the same pipeline.
# 1. $CONFIG_DIR/config.yaml - full config override, used verbatim
# 2. $CONFIG_DIR/scrape_configs.yaml - custom scrape_configs spliced into the template
# 3. SCRAPE_TARGETS env var - static scrape job rendered from env
# 4. $CONFIG_DIR/gcp_metrics_list.yaml - custom metrics_list for the GCM receiver
# 5. GCP_PROJECT_ID + GCP_METRICS / GCP_METRIC_FILTERS - GCM receiver rendered from env
#
# Scalar values (API key, host, interval, ...) are left as ${env:VAR} references
# for the collector's native config substitution, so secrets never pass through
# this script. Only the scrape_configs and metrics_list block structures are
# rendered here, because env substitution cannot expand a comma-separated
# string into a YAML list.
#
# RENDER_ONLY=1 prints the resolved config and exits (used by tests/render).
set -eu
CONFIG_DIR="${CONFIG_DIR:-/etc/posthog}"
TEMPLATE="$CONFIG_DIR/config.yaml.tmpl"
RENDERED="${RENDERED_CONFIG:-/tmp/config.yaml}"
OTELCOL="${OTELCOL_BIN:-/usr/local/bin/otelcol-contrib}"
finish() {
if [ "${RENDER_ONLY:-}" = "1" ]; then
cat "$1"
exit 0
fi
exec "$OTELCOL" --config "$1"
}
# Sharding: with SHARD_COUNT set, each instance keeps only the targets that
# hash to its SHARD_INDEX, so a fleet partitions the target set with no
# coordination. The index falls back to the hostname's trailing ordinal
# (StatefulSet pods are named <name>-<ordinal>), matching how vmagent shards.
# Resolved before any config path so mounted configs can reference
# ${env:SHARD_INDEX} too.
SHARD_COUNT="${SHARD_COUNT:-1}"
case "$SHARD_COUNT" in
'' | *[!0-9]*)
echo "error: SHARD_COUNT must be a positive integer, got '$SHARD_COUNT'" >&2
exit 1
;;
esac
if [ "$SHARD_COUNT" -gt 1 ]; then
if [ -z "${SHARD_INDEX:-}" ]; then
host="${HOSTNAME:-$(hostname)}"
SHARD_INDEX=$(printf '%s' "$host" | sed -n 's/.*-\([0-9][0-9]*\)$/\1/p')
if [ -z "$SHARD_INDEX" ]; then
echo "error: SHARD_INDEX is not set and cannot be derived from hostname '$host'" \
"(expected a trailing -<ordinal>, e.g. a StatefulSet pod name)" >&2
exit 1
fi
fi
case "$SHARD_INDEX" in
'' | *[!0-9]*)
echo "error: SHARD_INDEX must be a non-negative integer, got '$SHARD_INDEX'" >&2
exit 1
;;
esac
if [ "$SHARD_INDEX" -ge "$SHARD_COUNT" ]; then
echo "error: SHARD_INDEX must be less than SHARD_COUNT ($SHARD_INDEX >= $SHARD_COUNT)" >&2
exit 1
fi
export SHARD_INDEX
fi
if [ -f "$CONFIG_DIR/config.yaml" ]; then
finish "$CONFIG_DIR/config.yaml"
fi
if [ -z "${POSTHOG_API_KEY:-}" ]; then
echo "error: POSTHOG_API_KEY is required (your PostHog project API key)" >&2
exit 1
fi
PROM_SNIPPET=$(mktemp)
GCP_SNIPPET=$(mktemp)
trap 'rm -f "$PROM_SNIPPET" "$GCP_SNIPPET"' EXIT
# Double single quotes so a value stays valid inside YAML single quotes.
yaml_squote() {
printf '%s' "$1" | sed "s/'/''/g"
}
# --- prometheus source ---
if [ -f "$CONFIG_DIR/scrape_configs.yaml" ]; then
# Re-indent the mounted scrape_configs list under the prometheus receiver.
{
printf '%s\n' ' prometheus:' ' config:' ' scrape_configs:'
awk '{ if ($0 == "") print ""; else print " " $0 }' \
"$CONFIG_DIR/scrape_configs.yaml"
} >"$PROM_SNIPPET"
elif [ -n "${SCRAPE_TARGETS:-}" ]; then
TARGETS=""
OLDIFS=$IFS
IFS=,
for target in $SCRAPE_TARGETS; do
target=$(printf '%s' "$target" | sed 's/^ *//;s/ *$//')
[ -n "$target" ] || continue
target=$(yaml_squote "$target")
if [ -n "$TARGETS" ]; then
TARGETS="$TARGETS, '$target'"
else
TARGETS="'$target'"
fi
done
IFS=$OLDIFS
if [ -z "$TARGETS" ]; then
echo "error: SCRAPE_TARGETS contained no targets" >&2
exit 1
fi
cat >"$PROM_SNIPPET" <<EOF
prometheus:
config:
scrape_configs:
- job_name: '\${env:SCRAPE_JOB_NAME:-posthog-metrics-agent}'
scrape_interval: '\${env:SCRAPE_INTERVAL:-15s}'
metrics_path: '\${env:SCRAPE_METRICS_PATH:-/metrics}'
# OpenMetrics first so exemplars (trace links) survive the scrape.
scrape_protocols: [OpenMetricsText1.0.0, OpenMetricsText0.0.1, PrometheusText0.0.4]
static_configs:
- targets: [$TARGETS]
EOF
if [ "$SHARD_COUNT" -gt 1 ]; then
cat >>"$PROM_SNIPPET" <<EOF
# Shard $SHARD_INDEX of $SHARD_COUNT: keep only targets that hash to this shard.
relabel_configs:
- source_labels: [__address__]
modulus: $SHARD_COUNT
target_label: __tmp_shard
action: hashmod
- source_labels: [__tmp_shard]
regex: '$SHARD_INDEX'
action: keep
EOF
fi
fi
PROM_ENABLED=0
[ -s "$PROM_SNIPPET" ] && PROM_ENABLED=1
# --- google cloud monitoring source ---
GCP_ENABLED=0
if [ -n "${GCP_PROJECT_ID:-}" ]; then
GCP_ENABLED=1
fi
if [ "$GCP_ENABLED" -eq 0 ] && {
[ -n "${GCP_METRICS:-}" ] || [ -n "${GCP_METRIC_FILTERS:-}" ] || [ -f "$CONFIG_DIR/gcp_metrics_list.yaml" ]
}; then
echo "error: GCP_PROJECT_ID is required when GCP_METRICS or GCP_METRIC_FILTERS is set" \
"or $CONFIG_DIR/gcp_metrics_list.yaml is mounted" >&2
exit 1
fi
if [ "$GCP_ENABLED" -eq 1 ]; then
if [ "$SHARD_COUNT" -gt 1 ]; then
echo "error: GCP_PROJECT_ID cannot be combined with SHARD_COUNT > 1" \
"(every shard would pull the same Cloud Monitoring series);" \
"run a separate single-instance agent for Google Cloud Monitoring" >&2
exit 1
fi
if [ -n "${GOOGLE_APPLICATION_CREDENTIALS:-}" ] && [ ! -r "$GOOGLE_APPLICATION_CREDENTIALS" ]; then
echo "error: GOOGLE_APPLICATION_CREDENTIALS points to '$GOOGLE_APPLICATION_CREDENTIALS'" \
"which is missing or not readable by uid 10001" >&2
exit 1
fi
cat >"$GCP_SNIPPET" <<EOF
googlecloudmonitoring:
project_id: '\${env:GCP_PROJECT_ID}'
collection_interval: '\${env:GCP_COLLECTION_INTERVAL:-60s}'
metrics_list:
EOF
if [ -f "$CONFIG_DIR/gcp_metrics_list.yaml" ]; then
# Re-indent the mounted metrics_list under the receiver.
awk '{ if ($0 == "") print ""; else print " " $0 }' \
"$CONFIG_DIR/gcp_metrics_list.yaml" >>"$GCP_SNIPPET"
else
if [ -n "${GCP_METRICS:-}" ]; then
OLDIFS=$IFS
IFS=,
for metric in $GCP_METRICS; do
metric=$(printf '%s' "$metric" | sed 's/^ *//;s/ *$//')
[ -n "$metric" ] || continue
printf " - metric_name: '%s'\n" "$(yaml_squote "$metric")" >>"$GCP_SNIPPET"
done
IFS=$OLDIFS
fi
if [ -n "${GCP_METRIC_FILTERS:-}" ]; then
# Filters can contain commas (one_of("a", "b")), so they split on
# semicolons, which never appear in the filter grammar.
OLDIFS=$IFS
IFS=';'
for filter in $GCP_METRIC_FILTERS; do
filter=$(printf '%s' "$filter" | sed 's/^ *//;s/ *$//')
[ -n "$filter" ] || continue
printf " - metric_descriptor_filter: '%s'\n" "$(yaml_squote "$filter")" >>"$GCP_SNIPPET"
done
IFS=$OLDIFS
fi
fi
if ! grep -q 'metric_' "$GCP_SNIPPET"; then
echo "error: GCP_METRICS or GCP_METRIC_FILTERS is required with GCP_PROJECT_ID" \
"(or mount $CONFIG_DIR/gcp_metrics_list.yaml with at least one entry)" >&2
exit 1
fi
fi
if [ "$PROM_ENABLED" -eq 0 ] && [ "$GCP_ENABLED" -eq 0 ]; then
echo "error: SCRAPE_TARGETS is required (comma-separated host:port list)," \
"unless you mount $CONFIG_DIR/scrape_configs.yaml or $CONFIG_DIR/config.yaml," \
"or set GCP_PROJECT_ID to pull from Google Cloud Monitoring" >&2
exit 1
fi
DEBUG_ENABLED=0
case "${POSTHOG_DEBUG:-}" in
1 | true | TRUE | yes) DEBUG_ENABLED=1 ;;
esac
PERSIST_ENABLED=0
case "${PERSIST_QUEUE:-}" in
1 | true | TRUE | yes) PERSIST_ENABLED=1 ;;
esac
awk -v prom="$PROM_SNIPPET" -v gcp="$GCP_SNIPPET" -v prom_on="$PROM_ENABLED" -v gcp_on="$GCP_ENABLED" \
-v debug="$DEBUG_ENABLED" -v persist="$PERSIST_ENABLED" '
function splice(file) {
while ((getline line < file) > 0) print line
close(file)
}
$0 == "#__PROMETHEUS_RECEIVER__" {
if (prom_on == "1") splice(prom)
next
}
$0 == "#__GCP_RECEIVER__" {
if (gcp_on == "1") splice(gcp)
next
}
$0 == "#__GCP_RESOURCE_PROCESSOR__" {
if (gcp_on == "1") {
print " # Cloud Monitoring resources carry no service.name; give them one so they"
print " # group in the Metrics UI. `insert` never overrides a scraped job_name."
print " resource/gcp:"
print " attributes:"
print " - key: service.name"
print " value: \x27${env:GCP_SERVICE_NAME:-google-cloud-monitoring}\x27"
print " action: insert"
}
next
}
$0 == "#__DEBUG_EXPORTER__" {
if (debug == "1") {
print " debug:"
print " verbosity: detailed"
}
next
}
$0 == "#__SENDING_QUEUE__" {
if (persist == "1") {
print " sending_queue:"
print " enabled: true"
print " # Survives restarts: batches persist to disk until delivered."
print " storage: file_storage"
}
next
}
$0 == "#__FILE_STORAGE__" {
if (persist == "1") {
print " file_storage:"
print " directory: \x27${env:QUEUE_DIR:-/var/lib/posthog-agent}\x27"
print " create_directory: true"
}
next
}
index($0, "__PIPELINE_RECEIVERS__") {
if (prom_on == "1" && gcp_on == "1") {
receivers = "prometheus, googlecloudmonitoring"
} else if (gcp_on == "1") {
receivers = "googlecloudmonitoring"
} else {
receivers = "prometheus"
}
sub(/__PIPELINE_RECEIVERS__/, receivers)
print
next
}
index($0, "__PIPELINE_PROCESSORS__") {
processors = (gcp_on == "1") ? "memory_limiter, resource/gcp, batch" : "memory_limiter, batch"
sub(/__PIPELINE_PROCESSORS__/, processors)
print
next
}
index($0, "__SERVICE_EXTENSIONS__") {
extensions = (persist == "1") ? "health_check, file_storage" : "health_check"
sub(/__SERVICE_EXTENSIONS__/, extensions)
print
next
}
index($0, "__PIPELINE_EXPORTERS__") {
exporters = (debug == "1") ? "otlphttp, debug" : "otlphttp"
sub(/__PIPELINE_EXPORTERS__/, exporters)
print
next
}
{ print }
' "$TEMPLATE" >"$RENDERED"
finish "$RENDERED"