[nats helm 1.x] nats-server pid location changes based on configuration

[mprimi]

Using Helm chart nats-1.0.0-beta6

With:

  reloader:
    enabled: false

The nats-server pid file is found at: /var/run/nats.pid

With:

  reloader:
    enabled: true

The nats-server pid file is found at a different location: /var/run/nats/nats.pid

[caleblloyd]

This is by design. When the reloaded is enabled, the PID file needs to be accessible by both the nats and reloader containers.

When the reloader is enabled, an emptyDir volume is created at /var/run/nats and mounted to both containers, and the PID file is written inside of that volume

When the reloader is disabled, this volume does not exit, and the /var/run/nats directory does not exist, so the pidfile goes in /var/run instead.

0.x may have always been making this volume. But it is not necessary to make when the reloader is not enabled so 1.x changes to only make the volume if it is needed.

[MikaelElkiaer]

@caleblloyd Sorry to piggyback on this issue.
But I suspect that I am facing a similar issue - related to config and resulting pid set-up.

I have reloader disabled, but am running all containers as non-root.
This caused an error when starting the server - stating the server did not have permissions to write to /var/run/nats.pid.

As a temporary fix, I added to my config:

container:
  patch:
    - op: add
      path: /volumeMounts/-
      value:
        mountPath: /var/run
        name: var-run
podTemplate:
  patch:
      - op: add
        path: /spec/volumes/-
        value:
          emptyDir: {}
          name: var-run

[caleblloyd]

Ah, I had not considered the unprivileged use case. Seems that it is best to always make the emptyDir then - #782