kubeapps-postgresql-0 CrashLoopBackOff

[sjxuereb]

Describe the bug

Postgres keeps crashing after installation. I tried with a standard installation and also with an nfs storageclass. it keeps crashing on the simplest of installations/configuration. I tried installing kubeapps several times.

After logging in kubeapps dashboard the catalog takes a long time to load and then I notice the postgres pod crashes. The pod revives and crashes on every dashboard command.

kubeapps-postgresql-0 0/1 CrashLoopBackOff

OS: Ubuntu 24.04.1 LTS
K8: Client Version: v1.31.3, Kustomize Version: v5.4.2, Server Version: v1.31.3
Helm: Version:"v3.16.3"

CHART NAME: kubeapps
CHART VERSION: 17.0.3
APP VERSION: 2.12.0**

To Reproduce

helm install kubeapps oci://registry-1.docker.io/bitnamicharts/kubeapps --namespace kapps --set useHelm3=true --set postgresql.resourcesPreset=2xlarge

Control 10g Ram /8 cores/ 32G disk
Workers 8g Ram /4cores/ 32G disk

Expected behavior

Normal behavior of the postgres and kubeapps dashboard.

Screenshots

Additional context
I also tried with a storageclass. Same results, postgres keeps crashing.

--set global.defaultStorageClass=nfs-client
--set postgresql.primary.persistence.enabled=true
--set redis.master.persistence.enabled=true
--set redis.replica.persistence.enabled=true

[sjxuereb]

I have since re-installed the chart as:

helm install kubeapps oci://registry-1.docker.io/bitnamicharts/kubeapps -n kubeapps
--set global.defaultStorageClass=nfs-client
--set frontend.service.type=LoadBalancer
--set postgresql.primary.persistence.enabled=true
--set postgresql.auth.postgresPassword=password
--set postgresql.resourcesPreset=medium
--set redis.master.persistence.enabled=true
--set redis.auth.password=password
--set redis.master.resourcesPreset=medium

The data folder has been created on the nfs and postgres runs for a few minutes. It then dies out of memory.

journalctl -k | grep 'Killed process' ( on the worker node gives ) :
Memory cgroup out of memory: Killed process 32350 (postgres)

As an extra test I installed the postgres chart on its own and it worked with no issues. However the kubeapps embedded chart is failing with OOMKilled
https://github.com/bitnami/charts/tree/main/bitnami/postgresql#parameters

Any help please. Thanks

[singatias]

I had the same issue following the guide to test it out

[fea359]

Changing the StatefulSet of postgresql and increasing the upper limits of CPU and memory resources, for example, changing the CPU to 2 and the memory to 4096m, or directly changing to unlimited, can solve the problem.

[beltran-rubo]

That looks like a very specific issue for NFS storage volume. Depending on the performance of the dis, it is possible in certain scenarios require to increase the CPU & memory for postgres.

[maro-kt]

I am also encountering the same issue. Upon investigation, it seems that Helm's variables are not being properly reflected. Even when setting the values for postgresql.resources or postgresql.resourcesPreset , they do not appear in the actual manifest.

$ helm upgrade -n kubeapps kubeapps bitnami/kubeapps --set postgresql.resourcesPreset=medium
$ helm get values kubeapps -n kubeapps 

USER-SUPPLIED VALUES:
postgresql:
  resourcesPreset: medium

Execution result:

$ helm get manifest kubeapps -n kubeapps 

# Source: kubeapps/charts/postgresql/templates/primary/statefulset.yaml
apiVersion: apps/v1
kind: StatefulSet
metadata:
（omitted for brevity）
          resources:
            limits:
              cpu: 150m
              ephemeral-storage: 2Gi
              memory: 192Mi
            requests:
              cpu: 100m
              ephemeral-storage: 50Mi
              memory: 128Mi
          volumeMounts:
            - name: empty-dir
              mountPath: /tmp
              subPath: tmp-dir
            - name: empty-dir
              mountPath: /opt/bitnami/postgresql/conf
              subPath: app-conf-dir
            - name: empty-dir
              mountPath: /opt/bitnami/postgresql/tmp
              subPath: app-tmp-dir
            - name: dshm
              mountPath: /dev/shm
            - name: data
              mountPath: /bitnami/postgresql
      volumes:
        - name: empty-dir
          emptyDir: {}
        - name: dshm
          emptyDir:
            medium: Memory
        - name: data
          emptyDir: {}