Add a hand-written PostgreSQL chart for toolshed

Not Bitnami's: that registry has been actively unstable here (it broke
Contour twice, infra issue #4) and PostgreSQL publishes no official chart.
A single StatefulSet, PVC and Service is small enough that owning it costs
less than depending on an unstable repackage.

Credentials come from an existing Secret rather than being generated by
the chart — a chart that generates its own password regenerates it on
every render and silently locks you out of the existing volume.

Details that matter and are easy to get wrong:
  - PGDATA is a subdirectory of the mount, not the mount itself. initdb
    refuses to run in a directory that already has contents.
  - Probes run through a shell. Kubernetes does not expand $(VAR) inside
    exec probe commands, only in command/args.
  - fsGroup 70 so the volume stays writable after the entrypoint drops
    from root to the postgres user on the Alpine variant.
  - shared_buffers cut to 32MB from PostgreSQL's 128MB default. The node
    has 8GB and was at its ceiling before this.

Verified with helm template.
This commit is contained in:
Mukul Sharma
2026-09-04 16:21:26 +05:30
parent 5d4d3d44e5
commit d4bb8985fb
5 changed files with 223 additions and 0 deletions
+18
View File
@@ -0,0 +1,18 @@
apiVersion: v2
name: postgresql
description: |
Single-instance PostgreSQL for this homelab.
Hand-written rather than vendoring Bitnami's chart: Broadcom has been
retiring and freezing images behind that repo (see claude.md infra issue
#4, where it broke Contour twice), and PostgreSQL publishes no official
Helm chart of its own. A single StatefulSet with one PVC is small enough
that owning it outright costs less than depending on an unstable
repackage.
Not highly available and not intended to be. One replica, one PVC, no
replication, no connection pooler. Adding any of those to a single-node
cluster would be theatre.
type: application
version: 0.1.0
appVersion: "16"
@@ -0,0 +1,23 @@
{{- $name := .Values.fullnameOverride | default "postgresql" -}}
# ClusterIP only. Nothing outside the cluster should reach the database, and
# there is no Ingress here on purpose — Contour terminates HTTP, and exposing
# PostgreSQL's wire protocol through it is neither possible nor wanted.
#
# Consumers address this as:
# {{ $name }}.{{ .Release.Namespace }}.svc.cluster.local:{{ .Values.service.port }}
apiVersion: v1
kind: Service
metadata:
name: {{ $name }}
namespace: {{ .Release.Namespace }}
labels:
app: {{ $name }}
spec:
type: ClusterIP
selector:
app: {{ $name }}
ports:
- name: postgres
port: {{ .Values.service.port }}
targetPort: postgres
protocol: TCP
@@ -0,0 +1,89 @@
{{- $name := .Values.fullnameOverride | default "postgresql" -}}
apiVersion: apps/v1
kind: StatefulSet
metadata:
name: {{ $name }}
namespace: {{ .Release.Namespace }}
labels:
app: {{ $name }}
spec:
serviceName: {{ $name }}
replicas: 1
selector:
matchLabels:
app: {{ $name }}
template:
metadata:
labels:
app: {{ $name }}
spec:
securityContext:
# The official image starts as root, initialises the data directory,
# then drops to the postgres user (uid 70 on the Alpine variant).
# fsGroup makes the provisioned volume group-writable so that drop
# still leaves the data directory usable.
fsGroup: 70
terminationGracePeriodSeconds: 60
containers:
- name: postgres
image: "{{ .Values.image.repository }}:{{ .Values.image.tag }}"
imagePullPolicy: {{ .Values.image.pullPolicy }}
args:
- -c
- shared_buffers={{ .Values.config.sharedBuffers }}
- -c
- max_connections={{ .Values.config.maxConnections }}
env:
- name: POSTGRES_USER
valueFrom:
secretKeyRef:
name: {{ .Values.existingSecret }}
key: {{ .Values.secretKeys.username }}
- name: POSTGRES_PASSWORD
valueFrom:
secretKeyRef:
name: {{ .Values.existingSecret }}
key: {{ .Values.secretKeys.password }}
- name: POSTGRES_DB
value: {{ .Values.database | quote }}
# PGDATA must be a SUBDIRECTORY of the mount, not the mount
# itself: initdb refuses to run in a directory that already has
# contents, and a freshly provisioned volume is not always empty.
- name: PGDATA
value: /var/lib/postgresql/data/pgdata
ports:
- name: postgres
containerPort: 5432
protocol: TCP
# Run through a shell so the environment expands — Kubernetes does
# not substitute $(VAR) inside exec probe commands.
readinessProbe:
exec:
command: ["sh", "-c", 'pg_isready -U "$POSTGRES_USER" -d "$POSTGRES_DB" -h 127.0.0.1']
initialDelaySeconds: 10
periodSeconds: 10
timeoutSeconds: 5
failureThreshold: 6
livenessProbe:
exec:
command: ["sh", "-c", 'pg_isready -U "$POSTGRES_USER" -d "$POSTGRES_DB" -h 127.0.0.1']
initialDelaySeconds: 30
periodSeconds: 20
timeoutSeconds: 5
failureThreshold: 6
resources:
{{- toYaml .Values.resources | nindent 12 }}
volumeMounts:
- name: data
mountPath: /var/lib/postgresql/data
{{- if .Values.persistence.enabled }}
volumeClaimTemplates:
- metadata:
name: data
spec:
accessModes: ["ReadWriteOnce"]
storageClassName: {{ .Values.persistence.storageClass | quote }}
resources:
requests:
storage: {{ .Values.persistence.size | quote }}
{{- end }}
+45
View File
@@ -0,0 +1,45 @@
# Chart defaults. Real configuration lives in
# helm-overrides/k8s-admin-prd-ase1/postgresql/custom-values.yaml.
fullnameOverride: postgresql
image:
# Pulled from Docker Hub, like every other infra component here (gitea,
# vault, harbor). The base-images mirror in Harbor exists to remove
# Docker Hub from the *application build* path and to minimise shipped
# app images — it is not in play for platform components.
repository: postgres
tag: "16-alpine"
pullPolicy: IfNotPresent
# Name of the Secret holding username/password. Created by External
# Secrets from Vault, not by this chart — a chart that generates its own
# database password regenerates it on every render, which silently locks
# you out of an existing volume.
existingSecret: postgresql-credentials
secretKeys:
username: username
password: password
database: toolshed
service:
port: 5432
persistence:
enabled: true
storageClass: local-path
size: 5Gi
# Tuned down hard. The box has 8GB total and is at its ceiling; PostgreSQL's
# default shared_buffers of 128MB is most of this pod's budget on its own.
config:
sharedBuffers: 32MB
maxConnections: "50"
resources:
requests:
cpu: 50m
memory: 64Mi
limits:
memory: 256Mi