When first starting Plausible from the chart, the liveness probe reliably kills the plausible-analytics deployment before it has a chance to complete its initial migrations and become ready.
Version of Helm, Kubernetes and the plausible-analytics chart:
$ kubectl version
Client Version: v1.29.1
Kustomize Version: v5.0.4-0.20230601165947-6ce0bf390ce3
Server Version: v1.28.5
$ helm version
version.BuildInfo{Version:"v3.13.2", GitCommit:"2a2fb3b98829f1e0be6fb18af2f6599e0f4e8243", GitTreeState:"clean", GoVersion:"go1.21.4"}
I used version 0.2.4 of the plausible-analytics chart
What happened:
Kubernetes kills Plausible when the logs reach this state:
Loading plausible..
Starting dependencies..
Starting repos..
create Plausible.Repo database if it doesn't exist
create Plausible.IngestRepo database if it doesn't exist
Creation of Db successful!
+ /entrypoint.sh db migrate
Loading plausible..
Starting dependencies..
Starting repos..
Running migrations for Elixir.Plausible.Repo
What you expected to happen:
A startup probe or a configurable initialDelaySeconds on the liveness probe allow the pod to take extra time to start up. Ideally, I can configure the probes in the values file.
How to reproduce it (as minimally and precisely as possible):
Have a slow server and start with an empty database. I pull the Maxmind GeoIP database and limit to 100m CPU, that may help reproducing this.
Describe the bug
When first starting Plausible from the chart, the liveness probe reliably kills the
plausible-analytics
deployment before it has a chance to complete its initial migrations and become ready.Version of Helm, Kubernetes and the plausible-analytics chart:
I used version 0.2.4 of the
plausible-analytics
chartWhat happened:
Kubernetes kills Plausible when the logs reach this state:
What you expected to happen:
A startup probe or a configurable
initialDelaySeconds
on the liveness probe allow the pod to take extra time to start up. Ideally, I can configure the probes in the values file.How to reproduce it (as minimally and precisely as possible):
Have a slow server and start with an empty database. I pull the Maxmind GeoIP database and limit to
100m
CPU, that may help reproducing this.