You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
nextcloud.openmetrics.allowedClients is silently ignored unless nextcloud.configs is set #887
nextcloud.openmetrics.allowedClients renders OPENMETRICS_ALLOWED_CLIENTS into the app
container, but nothing in the pod ever reads it. The only reader is files/defaultConfigs/helm-metrics.config.php.tpl, and the defaultConfigs files are only
mounted when nextcloud.configs is non-empty
(_helpers.tpl#L431); templates/config.yaml renders no ConfigMap at all otherwise.
So with the chart's default nextcloud.configs: {}, setting allowedClients does nothing:
Nextcloud keeps its built-in default of 127.0.0.1 only, and Prometheus is refused with 403.
This is the same gate reported in #761, but it bites differently here. The argument there for
keeping it — "normally you do not need the defaultConfig, because the files are already part of
the container image" — does not hold for helm-metrics.config.php. Like imaginary.config.php,
it is a chart file, not one of nextcloud/docker's.
There is no image copy to fall back on, so nextcloud.openmetrics.allowedClients cannot work on
its own under any configuration.
It also affects the chart's own defaults: prometheus.serviceMonitor selects the app Service as
well as the exporter Service, so an out-of-the-box metrics.enabled: true + prometheus.serviceMonitor.enabled: true install scrapes /metrics on the app pod, gets a 403,
and sits with a permanently down target firing TargetDown.
Steps to reproduce
nextcloud:
openmetrics:
allowedClients:
- "127.0.0.1"
- "10.244.0.0/16"# your pod CIDRmetrics:
enabled: trueprometheus:
serviceMonitor:
enabled: true
$ kubectl exec deploy/nextcloud -c nextcloud -- sh -c 'echo $OPENMETRICS_ALLOWED_CLIENTS'127.0.0.1,10.244.0.0/16
$ kubectl exec deploy/nextcloud -c nextcloud -- ls /var/www/html/config/apache-pretty-urls.config.php apcu.config.php apps.config.php autoconfig.phpconfig.php config.sample.php redis.config.php reverse-proxy.config.phps3.config.php smtp.config.php swift.config.php upgrade-disable-web.config.php
# no helm-metrics.config.php
$ kubectl exec deploy/nextcloud -c nextcloud -- sh -c 'cd /var/www/html && php occ config:system:get openmetrics_allowed_clients'
# unset
$ # from a pod whose IP is inside the CIDR above
$ curl -o /dev/null -w '%{http_code}\n' http://<nextcloud-pod-ip>:80/metrics403
Adding any entry to nextcloud.configs makes it work, which is the tell.
Expected behaviour
Setting nextcloud.openmetrics.allowedClients configures openmetrics_allowed_clients, without
also having to set an unrelated value.
Possible fix
Drop the {{- if .Values.nextcloud.configs }} guard around the defaultConfigs loop in _helpers.tpl and around the ConfigMap in config.yaml, as #761 asks. At minimum, helm-metrics.config.php and imaginary.config.php need to mount whenever they are enabled,
since neither exists in the container image.
Activity
fix(nextcloud): apply the OpenMetrics client allowlist the chart igno…