Add Apache status monitoring to Helm chart
Publish PHP Apache image / Build, test and push (push) Successful in 1m10s

This commit is contained in:
feedc0de committed 2026-09-10 20:49:26 +02:00
1 parent 0396053c47
commit 5acc3b0154
11 files changed
+260 -7

No files matched your search

+20 -5
View File
@@ -13,7 +13,7 @@ upgrade also upgrades the runtime image.
Add and refresh the Gitea chart repository with:
```sh
helm repo add brunner https://code.brunner.ninja/api/packages/feedc0de/helm
helm repo add brunner https://brunner.ninja/charts
helm repo update brunner
```
@@ -40,6 +40,7 @@ helm upgrade --install example-site brunner/php-homepage \
--namespace default \
--values /path/to/example-site/values.yaml \
--take-ownership \
--force-conflicts \
--wait
```
@@ -55,7 +56,21 @@ With `persistence.existingClaim` empty, the chart creates an RWX PVC using
uninstall does not discard site content. Set `persistence.retain: false` only
when deleting the release should also delete its dynamically created claim.
Ingress, TLS, probes, scheduling, storage size, and the image tag are all
configurable through `values.yaml`. TCP probes are used by default because
individual websites may redirect `/`, require authentication, or select their
own HTTP status behavior.
Ingress, TLS, probes, scheduling, storage size, Apache's `ServerAdmin`, and the
image tag are all configurable through `values.yaml`. TCP startup, readiness,
and liveness probes are enabled by default because individual websites may
redirect `/`, require authentication, or select their own HTTP status behavior.
## Apache metrics
The image provides an extended `mod_status` endpoint on pod-local loopback port
8080. It is deliberately not exposed directly by the Service because extended
status can contain current client and request details.
Set `metrics.enabled` to add an Apache exporter sidecar and expose its sanitized
Prometheus metrics on the Service. `metrics.serviceMonitor.enabled` and
`metrics.prometheusRule.enabled` render the corresponding Prometheus Operator
resources when their CRDs exist. The rules cover exporter scrape failures,
Apache status failures, and worker saturation. No Prometheus namespace, release
label, or Alertmanager is hard-coded; selection and alert routing remain the
responsibility of the cluster's Prometheus installation.