[uptime-kuma] Support per-ingress probe path annotation
## Context The `external-monitor-sync` CronJob probed `https://<host>/` for every `*.viktorbarzin.me` ingress. Homepages frequently return 200 (or allow-listed 30x/40x) even when the backend or DB is broken, producing false-negatives — the forgejo outage on 2026-04-17 was not caught for this reason: `/` returned a login page while `/api/healthz` returned 503 from the DB probe. Manual monitor edits don't stick: the next sync is create-if-missing only, so a deleted monitor gets recreated pointing at `/` again. ## This change Teaches the sync three things: 1. **Reads a new annotation** `uptime.viktorbarzin.me/external-monitor-path`. The annotation value is appended as the probe path; default `/` preserves today's behaviour for every ingress that hasn't opted in. 2. **Tightens accepted status codes** when an explicit path is set: `['200-299']` (strict — we expect a real healthz). The default `/` path keeps the existing lenient set `['200-299','300-399','400-499']` because homepages routinely 30x redirect or 40x on missing auth. 3. **Updates existing monitors** when the target URL or accepted status codes drift. Previously the loop was create-if-missing only, so annotating an already-monitored ingress had no effect until the monitor was deleted. Now re-running the sync after changing the annotation converges the live monitor. ## What is NOT in this change - No change to the Ingress annotations on any individual stack. Each service that wants a non-`/` probe path opts in separately. - No change to the ConfigMap fallback payload shape — legacy entries still get the lenient status codes. - Monitor DB state in Uptime Kuma's SQLite is untouched at plan time; the sync CronJob is what reconciles state on each run. ## Flow ``` ingress annotation CronJob Python ------------------ -------------- (none) --> url = https://host/ codes = lenient external-monitor-path --> url = https://host<path> codes = strict ['200-299'] ^^ "/api/healthz" https://host/api/healthz codes = ['200-299'] existing monitor + drifted target url --> api.edit_monitor(id, url=..., accepted_statuscodes=...) ``` ## Test Plan ### Automated - `terraform fmt -check -recursive stacks/uptime-kuma` — exit 0. - `scripts/tg plan` on `stacks/uptime-kuma` — `Plan: 0 to add, 1 to change, 0 to destroy`. The single in-place change is the CronJob command (Python heredoc re-rendered). No other resources drift. - Embedded Python compiles: extracted the `PYEOF` block and ran `python3 -m py_compile` — OK. ### Manual Verification 1. Annotate an ingress: `kubectl annotate ingress/<name> -n <ns> uptime.viktorbarzin.me/external-monitor-path=/api/healthz` 2. Trigger sync early: `kubectl -n uptime-kuma create job --from=cronjob/external-monitor-sync external-monitor-sync-manual` 3. Expected log line: `Updating monitor [External] <name>: https://host/ -> https://host/api/healthz (codes ['200-299','300-399','400-499'] -> ['200-299'])` 4. Inspect monitor in Uptime Kuma UI: URL and accepted status codes reflect the annotation. 5. Final summary line includes updated count: `Sync complete: 0 created, 1 updated, 0 deleted, N unchanged`. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Viktor Barzin
parent
842646ea4f
commit
f6812fe69f
1 changed files with 37 additions and 4 deletions
|
|
@ -339,6 +339,12 @@ FALLBACK_FILE = "/config/targets.json"
|
|||
PREFIX = "[External] "
|
||||
ANNOTATION_ENABLE = "uptime.viktorbarzin.me/external-monitor"
|
||||
ANNOTATION_NAME = "uptime.viktorbarzin.me/external-monitor-name"
|
||||
ANNOTATION_PATH = "uptime.viktorbarzin.me/external-monitor-path"
|
||||
DEFAULT_PATH = "/"
|
||||
# Homepages often serve 200/30x/40x even when backends are degraded.
|
||||
# When an explicit probe path is set we expect a real healthz: tighten codes.
|
||||
STATUSCODES_LENIENT = ["200-299", "300-399", "400-499"]
|
||||
STATUSCODES_STRICT = ["200-299"]
|
||||
SA_DIR = "/var/run/secrets/kubernetes.io/serviceaccount"
|
||||
API_SERVER = f"https://{os.environ.get('KUBERNETES_SERVICE_HOST', 'kubernetes.default.svc.cluster.local')}:{os.environ.get('KUBERNETES_SERVICE_PORT', '443')}"
|
||||
|
||||
|
|
@ -378,7 +384,11 @@ def load_from_api():
|
|||
if monitor_name in seen:
|
||||
continue # dedupe by final monitor name, not hostname (fixes duplicate creation)
|
||||
seen.add(monitor_name)
|
||||
targets.append({"name": label, "url": f"https://{host}"})
|
||||
path = anns.get(ANNOTATION_PATH) or DEFAULT_PATH
|
||||
if not path.startswith("/"):
|
||||
path = "/" + path
|
||||
statuscodes = STATUSCODES_STRICT if path != DEFAULT_PATH else STATUSCODES_LENIENT
|
||||
targets.append({"name": label, "url": f"https://{host}{path}", "statuscodes": statuscodes})
|
||||
return targets
|
||||
|
||||
|
||||
|
|
@ -386,7 +396,7 @@ def load_from_configmap():
|
|||
"""Legacy fallback: read the ConfigMap list."""
|
||||
with open(FALLBACK_FILE) as f:
|
||||
raw = json.load(f)
|
||||
return [{"name": t["name"], "url": t["url"]} for t in raw]
|
||||
return [{"name": t["name"], "url": t["url"], "statuscodes": STATUSCODES_LENIENT} for t in raw]
|
||||
|
||||
|
||||
try:
|
||||
|
|
@ -413,10 +423,12 @@ for m in monitors:
|
|||
existing_external[m["name"]] = m
|
||||
|
||||
target_names = set()
|
||||
targets_by_name = {}
|
||||
created = 0
|
||||
for t in targets:
|
||||
monitor_name = f"{PREFIX}{t['name']}"
|
||||
target_names.add(monitor_name)
|
||||
targets_by_name[monitor_name] = t
|
||||
if monitor_name not in existing_external:
|
||||
print(f"Creating monitor: {monitor_name} -> {t['url']}")
|
||||
api.add_monitor(
|
||||
|
|
@ -425,11 +437,31 @@ for t in targets:
|
|||
url=t["url"],
|
||||
interval=300,
|
||||
maxretries=3,
|
||||
accepted_statuscodes=["200-299", "300-399", "400-499"],
|
||||
accepted_statuscodes=t["statuscodes"],
|
||||
)
|
||||
created += 1
|
||||
time.sleep(0.3)
|
||||
|
||||
# Update monitors whose target URL or accepted status codes drifted
|
||||
# (e.g., new probe-path annotation added on an existing ingress).
|
||||
updated = 0
|
||||
for monitor_name, t in targets_by_name.items():
|
||||
existing = existing_external.get(monitor_name)
|
||||
if not existing:
|
||||
continue
|
||||
current_url = existing.get("url")
|
||||
current_codes = existing.get("accepted_statuscodes") or []
|
||||
if current_url == t["url"] and current_codes == t["statuscodes"]:
|
||||
continue
|
||||
print(f"Updating monitor {monitor_name}: {current_url} -> {t['url']} (codes {current_codes} -> {t['statuscodes']})")
|
||||
api.edit_monitor(
|
||||
existing["id"],
|
||||
url=t["url"],
|
||||
accepted_statuscodes=t["statuscodes"],
|
||||
)
|
||||
updated += 1
|
||||
time.sleep(0.3)
|
||||
|
||||
# Remove monitors for services no longer in the list
|
||||
deleted = 0
|
||||
for name, m in existing_external.items():
|
||||
|
|
@ -440,7 +472,8 @@ for name, m in existing_external.items():
|
|||
time.sleep(0.3)
|
||||
|
||||
api.disconnect()
|
||||
print(f"Sync complete: {created} created, {deleted} deleted, {len(target_names) - created} unchanged")
|
||||
unchanged = len(target_names) - created - updated
|
||||
print(f"Sync complete: {created} created, {updated} updated, {deleted} deleted, {unchanged} unchanged")
|
||||
PYEOF
|
||||
EOT
|
||||
]
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue