ci-pipeline-health: daily sweep of the off-infra CI chain (ADR-0002)

Viktor asked to monitor the pipelines closely as builds move off-infra
(PRD infra#10). New aux stack: daily 07:30 UTC CronJob on the
claude-agent-service image running a deterministic shell sweep —
GitHub Actions failures/stuck runs across owned repos, Woodpecker
pipeline failures, GHA free-tier minutes burn. Healthy = one quiet
Slack line; issues = Slack alert + comment on infra#10. In-cluster
(not a cloud routine) because Vault + the Woodpecker token are
LAN-only. Secrets via ExternalSecret (github_pat deliberately, not the
ghcr_pull_token alias — a scoped packages-only rotation couldn't read
Actions runs).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Viktor Barzin 2026-06-12 20:45:28 +00:00
parent 98f1f7fc24
commit d02ca4f2db
3 changed files with 280 additions and 0 deletions

View file

@ -0,0 +1,112 @@
#!/bin/sh
# ci-pipeline-health — daily sweep of the off-infra CI chain (ADR-0002, PRD infra#10).
# Deterministic (no LLM): GitHub Actions runs + Woodpecker pipelines + GHA minutes.
# Healthy => one quiet Slack line. Issues => Slack alert + comment on infra#10.
# POSIX sh + curl + jq only (runs on the Alpine claude-agent-service image).
# Exit 0 = sweep ran (even with findings); exit 2 = the sweep itself errored,
# which surfaces through the existing CronJob-failure alerting.
GH_API="https://api.github.com"
WP_API="https://ci.viktorbarzin.me/api"
WP_UI="https://ci.viktorbarzin.me"
NOW_EPOCH=$(date -u +%s)
SINCE_EPOCH=$((NOW_EPOCH - 86400))
SINCE_ISO=$(date -u -d "@${SINCE_EPOCH}" +%Y-%m-%dT%H:%M:%SZ)
PUSH_CUTOFF=$(date -u -d "@$((NOW_EPOCH - 259200))" +%Y-%m-%dT%H:%M:%SZ)
ISSUES=$(mktemp)
NOTES=$(mktemp)
trap 'rm -f "$ISSUES" "$NOTES"' EXIT
gha_checked=0
wp_checked=0
sweep_errors=0
gh_get() { curl -sf --max-time 30 -H "Authorization: Bearer ${GITHUB_PAT}" -H "Accept: application/vnd.github+json" "$1"; }
wp_get() { curl -sf --max-time 30 -H "Authorization: Bearer ${WOODPECKER_API_TOKEN}" "$1"; }
# --- 1) GitHub Actions runs across owned repos with a recent push ---
repos=$(gh_get "${GH_API}/user/repos?affiliation=owner&sort=pushed&per_page=60" \
| jq -r --arg cutoff "$PUSH_CUTOFF" '.[] | select(.pushed_at >= $cutoff) | .full_name')
if [ $? -ne 0 ]; then
echo "sweep: failed to list GitHub repos" >>"$ISSUES"; sweep_errors=1; repos=""
fi
for repo in $repos; do
runs=$(gh_get "${GH_API}/repos/${repo}/actions/runs?created=%3E%3D${SINCE_ISO}&per_page=50")
if [ $? -ne 0 ]; then echo "sweep: failed to list runs for ${repo}" >>"$ISSUES"; sweep_errors=1; continue; fi
n=$(printf '%s' "$runs" | jq '.workflow_runs | length')
gha_checked=$((gha_checked + n))
printf '%s' "$runs" | jq -r '.workflow_runs[]
| select(.conclusion == "failure" or .conclusion == "timed_out" or .conclusion == "cancelled" or .conclusion == "action_required")
| "GHA: \(.repository.full_name) #\(.run_number) [\(.name)] \(.conclusion) \(.html_url)"' >>"$ISSUES"
printf '%s' "$runs" | jq -r --argjson now "$NOW_EPOCH" '.workflow_runs[]
| select(.status == "in_progress" or .status == "queued")
| select(($now - ((.run_started_at // .created_at) | fromdateiso8601)) > 7200)
| "GHA stuck >2h: \(.repository.full_name) #\(.run_number) [\(.name)] \(.status) \(.html_url)"' >>"$ISSUES"
done
# --- 2) Woodpecker pipelines (deploy chain) ---
wrepos=$(wp_get "${WP_API}/repos?perPage=100" | jq -r '.[] | select(.active == true) | "\(.id) \(.full_name)"')
if [ $? -ne 0 ]; then
echo "sweep: failed to list Woodpecker repos" >>"$ISSUES"; sweep_errors=1; wrepos=""
fi
printf '%s\n' "$wrepos" | while IFS=' ' read -r id name; do
[ -z "$id" ] && continue
pls=$(wp_get "${WP_API}/repos/${id}/pipelines?perPage=10")
if [ $? -ne 0 ]; then echo "sweep: failed pipelines for ${name}" >>"$ISSUES"; continue; fi
printf '%s' "$pls" | jq -r --argjson since "$SINCE_EPOCH" --arg name "$name" --arg ui "$WP_UI" --arg id "$id" '
[.[] | select(.created >= $since)][]
| select(.status == "failure" or .status == "error" or .status == "killed")
| "Woodpecker: \($name) #\(.number) \(.status) (\(.event)) \($ui)/repos/\($id)/pipeline/\(.number)"' >>"$ISSUES"
printf '%s' "$pls" | jq --argjson since "$SINCE_EPOCH" '[.[] | select(.created >= $since)] | length' >>"$NOTES.wpcount" 2>/dev/null || true
done
wp_checked=$(awk '{s+=$1} END {print s+0}' "$NOTES.wpcount" 2>/dev/null || echo 0)
rm -f "$NOTES.wpcount"
# --- 3) GHA minutes vs free tier ---
billing=$(gh_get "${GH_API}/users/ViktorBarzin/settings/billing/actions")
if [ $? -eq 0 ]; then
used=$(printf '%s' "$billing" | jq -r '.total_minutes_used')
included=$(printf '%s' "$billing" | jq -r '.included_minutes')
if [ "${included:-0}" -gt 0 ] 2>/dev/null; then
pct=$((used * 100 / included))
echo "GHA minutes: ${used}/${included} (${pct}%)" >>"$NOTES"
[ "$pct" -ge 75 ] && echo "GHA minutes at ${pct}% of the free tier (${used}/${included}) — check for runaway workflows or consider Pro" >>"$ISSUES"
fi
else
echo "minutes check unavailable" >>"$NOTES"
fi
# v1 scope (deliberate, not silent): Forgejo→GitHub mirror-gap detection (a
# Forgejo push that produced no GHA run) is NOT implemented yet — it needs the
# per-repo mirror inventory that lands with the offinfra-onboard rollout (#13+).
# --- Report ---
issue_count=$(grep -c . "$ISSUES" || true)
summary="ci-pipeline-health: checked ${gha_checked} GHA runs + ${wp_checked} Woodpecker pipelines (24h). $(tr '\n' '; ' <"$NOTES")"
if [ "$issue_count" -eq 0 ]; then
text=":white_check_mark: ${summary}"
else
text=":rotating_light: ci-pipeline-health: ${issue_count} issue(s)
$(sed 's/^/• /' "$ISSUES")
${summary}"
body="Daily CI sweep found ${issue_count} issue(s):
$(sed 's/^/- /' "$ISSUES")
_${summary}_"
printf '%s' "$body" | jq -Rs '{body: .}' \
| curl -sf --max-time 30 -X POST -H "Authorization: Bearer ${GITHUB_PAT}" \
-H "Accept: application/vnd.github+json" \
-d @- "${GH_API}/repos/ViktorBarzin/infra/issues/10/comments" >/dev/null \
|| { echo "sweep: failed to comment on infra#10"; sweep_errors=1; }
fi
printf '%s' "$text" | jq -Rs '{text: .}' \
| curl -sf --max-time 30 -X POST -H 'Content-Type: application/json' -d @- "$SLACK_WEBHOOK" >/dev/null \
|| { echo "sweep: failed to post to Slack"; sweep_errors=1; }
echo "$text"
[ "$sweep_errors" -ne 0 ] && exit 2
exit 0

View file

@ -0,0 +1,159 @@
# ci-pipeline-health daily sweep of the off-infra CI chain (ADR-0002).
#
# Viktor's standing instruction (2026-06-12): monitor the pipelines closely
# during/after the off-infra builds migration (PRD infra#10). Deterministic
# shell sweep (files/sweep.sh) on the claude-agent-service image: GitHub
# Actions failures/stuck runs across owned repos, Woodpecker pipeline
# failures, GHA free-tier minutes burn. Healthy => one quiet Slack line;
# issues => Slack alert + a comment on infra#10.
#
# Runs IN-CLUSTER (not a claude.ai cloud routine) because Vault and the
# Woodpecker token are LAN-only cloud agents can't reach them.
variable "schedule" {
type = string
# 07:30 UTC = 08:30 London in summer (07:30 in winter acceptable drift,
# CronJob schedules are UTC-only).
default = "30 7 * * *"
}
# Mirrors stacks/claude-agent-service image tag the image ships curl + jq.
variable "image_tag" {
type = string
default = "2fd7670d"
}
locals {
namespace = "ci-pipeline-health"
image = "forgejo.viktorbarzin.me/viktor/claude-agent-service:${var.image_tag}"
labels = {
app = "ci-pipeline-health"
}
}
resource "kubernetes_namespace" "ci_pipeline_health" {
metadata {
name = local.namespace
labels = {
tier = local.tiers.aux
}
}
}
# github_pat (NOT the ghcr_pull_token alias): the sweep reads Actions runs +
# billing on PRIVATE mirrors, which a future scoped read:packages rotation of
# the alias could not do. Blast radius = this single-CronJob namespace.
resource "kubernetes_manifest" "external_secret" {
manifest = {
apiVersion = "external-secrets.io/v1beta1"
kind = "ExternalSecret"
metadata = {
name = "ci-pipeline-health-creds"
namespace = kubernetes_namespace.ci_pipeline_health.metadata[0].name
}
spec = {
refreshInterval = "15m"
secretStoreRef = {
name = "vault-kv"
kind = "ClusterSecretStore"
}
target = {
name = "ci-pipeline-health-creds"
}
data = [
{
secretKey = "GITHUB_PAT"
remoteRef = { key = "viktor", property = "github_pat" }
},
{
secretKey = "WOODPECKER_API_TOKEN"
remoteRef = { key = "ci/global", property = "woodpecker_api_token" }
},
{
secretKey = "SLACK_WEBHOOK"
remoteRef = { key = "ci/global", property = "slack_webhook" }
},
]
}
}
}
resource "kubernetes_config_map" "sweep_script" {
metadata {
name = "ci-pipeline-health-sweep"
namespace = kubernetes_namespace.ci_pipeline_health.metadata[0].name
}
data = {
"sweep.sh" = file("${path.module}/files/sweep.sh")
}
}
resource "kubernetes_cron_job_v1" "sweep" {
metadata {
name = "ci-pipeline-health"
namespace = kubernetes_namespace.ci_pipeline_health.metadata[0].name
labels = local.labels
}
spec {
schedule = var.schedule
concurrency_policy = "Forbid"
successful_jobs_history_limit = 3
failed_jobs_history_limit = 3
job_template {
metadata {
labels = local.labels
}
spec {
backoff_limit = 1
active_deadline_seconds = 600
ttl_seconds_after_finished = 86400
template {
metadata {
labels = local.labels
}
spec {
restart_policy = "Never"
image_pull_secrets {
name = "registry-credentials"
}
container {
name = "sweep"
image = local.image
command = ["/bin/sh", "/scripts/sweep.sh"]
env_from {
secret_ref {
name = "ci-pipeline-health-creds"
}
}
volume_mount {
name = "sweep-script"
mount_path = "/scripts"
read_only = true
}
resources {
requests = {
cpu = "50m"
memory = "64Mi"
}
limits = {
memory = "128Mi"
}
}
}
volume {
name = "sweep-script"
config_map {
name = kubernetes_config_map.sweep_script.metadata[0].name
}
}
}
}
}
}
}
lifecycle {
# KYVERNO_LIFECYCLE_V1
ignore_changes = [spec[0].job_template[0].spec[0].template[0].spec[0].dns_config]
}
depends_on = [kubernetes_manifest.external_secret]
}

View file

@ -0,0 +1,9 @@
include "root" {
path = find_in_parent_folders()
}
# ExternalSecret hits ESO which needs to be alive when the manifest applies.
dependency "external_secrets" {
config_path = "../external-secrets"
skip_outputs = true
}