fix(deployment): run plugin container as the reposilite user
Helm / helm-lint (push) Successful in 8s
Helm / helm-unittest (push) Successful in 27s
Generate README / generate-parameters (push) Successful in 42s
Markdown linter / markdown-link-checker (push) Successful in 38s
Markdown linter / markdown-lint (push) Successful in 42s
Release / publish-chart (push) Successful in 1m16s

The reposilite entrypoint chowns /app to 977:977 whenever it starts as root. The plugin container downloaded
the jar as its own image user (101), so after the first start the file was owned by 977:977 with mode 0644.
Since the plugins emptyDir survives a container restart within the same pod, a restarted plugin container
could no longer overwrite the existing jar and aborted with "Permission denied", which left the pod in a
permanent Init:CrashLoopBackOff until the pod itself was recreated.

Run the plugin container as 977:977 by default and expose the security context as
deployment.pluginContainer.securityContext, so it can be aligned when PUID/PGID are overridden.

Co-authored-by: Copilot <copilot@github.com>
This commit is contained in:
2026-09-27 18:13:56 +02:00
co-authored by Copilot
parent b2b9ab19a0
commit 562001dc0b
4 changed files with 60 additions and 42 deletions
+9
View File
@@ -178,6 +178,15 @@ deployment:
tag: "8.22.0"
pullPolicy: IfNotPresent
## @param deployment.pluginContainer.securityContext.runAsGroup Group id of the plugin container.
## @param deployment.pluginContainer.securityContext.runAsUser User id of the plugin container.
# Reposilite chowns /app to 977:977 when its entrypoint starts as root. Downloading the plugin as a
# different user leaves behind a file the plugin container can no longer overwrite, which breaks every
# restart that reuses the emptyDir. Align this with PUID/PGID when those are overridden.
securityContext:
runAsGroup: 977
runAsUser: 977
## @param deployment.priorityClassName PriorityClassName of the Reposilite deployment.
priorityClassName: ""