[K8S-STAND-PGSQL] Bad DB_CONNECTION_URI generated

[AurelienConte]

Hi,

I was installing the infisical standalone postgresql helm charts but nothing is starting up because of the DB_CONNECTION_URI badly generated.

Here is the line :

infisical/helm-charts/infisical-standalone-postgres/templates/_helpers.tpl

Lines 60 to 66 in 0042a95

	{{- define "infisical.postgresService" -}}
	{{- if .Values.postgresql.fullnameOverride -}}
	{{- .Values.postgresql.fullnameOverride | trunc 63 | trimSuffix "-" -}}
	{{- else -}}
	{{- printf "%s-postgresql" .Release.Name | trunc 63 | trimSuffix "-" -}}
	{{- end -}}
	{{- end -}}

infisical/helm-charts/infisical-standalone-postgres/templates/_helpers.tpl

Lines 72 to 73 in 0042a95

	{{- $serviceName := include "infisical.postgresService" . -}}
	{{- printf "postgresql://%s:%s@%s:5432/%s" $dbUsername $dbPassword $serviceName $dbName -}}

The default value of DB_CONNECTION_URI is : postgresql://infisical:root@postgresql:5432/infisicalDB

But in a kubernetes environment this can not work, a good string should use the ".svc.cluster.local" internal dns.

Assuming the charts is deployed in a namespace : "infisical" the DB_CONNECTION_URI should be :

postgresql://infisical:root@postgresql.infisical.svc.cluster.local:5432/infisicalDB

If you need more information on how to reproduce my problem feel free to ask me anything. ( But i was just following the guide : https://infisical.com/docs/self-hosting/deployment-options/kubernetes-helm )

Here is my values.yaml

infisical:
  image:
    tag: "v0.51.2-postgres"

ingress:
  enabled: false

[Grraahaam]

Hey @AurelienConte ! It's indeed a good practice to define the entire FQDN if ever you move services between namespaces (doesnt happen that often) or just to be sure to point to the right one.

Although, the kubelet automatically configure a /etc/resolf.conf in each pod for it to be able to resolve any hosts in the same namespace, so the current helm template works as expected.

$ cat /etc/resolv.conf 
search my-service.svc.cluster.local svc.cluster.local cluster.local my-namespace
nameserver 10.32.0.10
options ndots:5

If you want to point to a different postgresql instance (managed externally) you can configure it in the DB_CONNECTION_URI variable in the infisical-secrets k8s secret used by the infisical pod to get default/extra configs.

However, I don't understand why your instance isn't starting, it should.. Do you have any useful logs? Are you sure it's related to the DB_CONNECTION_URI?

Hope it helps

[AurelienConte]

Hi back!

I think I can say "My bad" because today after re-installing the chart there were no problems at all ...
I don't know why my instances couldn't connect to the Postgresql included in the helm chart.

I was not aware that the /etc/resolv.conf was filled automatically

We can close this issue since it was more a setup problem than a chart problem.

Thanks for your time!