Files
launchpad/apps/overlays/upc-dev/forte-drop-postgresql/resources/pg-backup-cronjob.yaml
T
29624e845d
/ test (push) Successful in 10s
fix(forte-drop-pg-backup): set MC_CONFIG_DIR so backups can upload (#23)
The nightly Postgres backup CronJob has been **failing every run** — no backups exist in `s3://drops/_pgbackups/`.

**Cause:** the upload container runs as uid 65532 (`runAsNonRoot`). `mc` defaults its config to `$HOME/.mc` = `/.mc` and dies with `mkdir /.mc: permission denied` on the non-writable root fs — before any upload.

**Fix:** set `MC_CONFIG_DIR=/work/.mc` (the shared emptyDir, writable via `fsGroup: 65532`). The `pg_dump` initContainer already succeeds; this lets the upload step actually run.

Validated: `kubectl kustomize` renders clean; env present on the upload container.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Danijel Simeunovic <danijel.simeunovic@fortedigital.com>
Co-authored-by: Sten <sten@Sten-sin-MacBook-Pro.local>
Reviewed-on: #23
Reviewed-by: Danijel Simeunovic <danijel.simeunovic@fortedigital.com>
Co-authored-by: Jørgen Stensrud <jorgen.stensrud@fortedigital.com>
Co-committed-by: Jørgen Stensrud <jorgen.stensrud@fortedigital.com>
2026-08-25 09:44:02 +00:00

100 lines
3.7 KiB
YAML

# Nightly logical backup of the forte-drop Postgres → UpCloud Managed Object Storage.
# Dumps to s3://drops/_pgbackups/ (the `_` prefix is collision-proof: app slugs match
# /^[a-z0-9][a-z0-9-]{0,62}$/ and can never start with `_`). Retains 30 days.
#
# Pod shape: initContainer pg_dump → shared emptyDir → mc upload + retention prune.
# Both images pinned. S3 creds reuse forte-drop-secrets (the app's UpCloud user has
# s3:* on the drops bucket). PG creds from forte-drop-pg-creds.
apiVersion: batch/v1
kind: CronJob
metadata:
name: forte-drop-pg-backup
namespace: forte-drop
labels:
app.kubernetes.io/name: postgresql
app.kubernetes.io/instance: forte-drop
app.kubernetes.io/component: backup
spec:
schedule: "0 2 * * *" # 02:00 UTC daily
concurrencyPolicy: Forbid
successfulJobsHistoryLimit: 3
failedJobsHistoryLimit: 3
jobTemplate:
spec:
backoffLimit: 2
template:
metadata:
labels:
app.kubernetes.io/name: postgresql
app.kubernetes.io/instance: forte-drop
app.kubernetes.io/component: backup
spec:
restartPolicy: Never
securityContext:
runAsNonRoot: true
runAsUser: 65532
fsGroup: 65532
volumes:
- name: work
emptyDir: {}
initContainers:
- name: dump
image: postgres:16-alpine
command:
- sh
- -c
- |
set -euo pipefail
TS=$(date -u +%Y%m%dT%H%M%SZ)
echo "dumping to /work/forte-drop-${TS}.sql.gz"
PGPASSWORD="$PGPASSWORD" pg_dump \
-h forte-drop-postgresql.forte-drop.svc \
-p 5432 -U "$PGUSER" -d drops \
--no-owner --no-privileges \
| gzip -9 > "/work/forte-drop-${TS}.sql.gz"
echo "dump complete: $(ls -lh /work/)"
env:
- name: PGUSER
valueFrom:
secretKeyRef: { name: forte-drop-pg-creds, key: pgusername }
- name: PGPASSWORD
valueFrom:
secretKeyRef: { name: forte-drop-pg-creds, key: pgpassword }
volumeMounts:
- name: work
mountPath: /work
containers:
- name: upload
image: quay.io/minio/mc:RELEASE.2024-11-21T17-21-54Z
command:
- sh
- -c
- |
set -euo pipefail
mc alias set obj "$S3_ENDPOINT" "$S3_KEY" "$S3_SECRET"
mc cp /work/*.sql.gz "obj/${S3_BUCKET}/_pgbackups/"
echo "uploaded. pruning backups older than 30d:"
mc rm --recursive --force --older-than 30d "obj/${S3_BUCKET}/_pgbackups/" || true
echo "backup retention pass complete"
env:
# mc writes its config under $MC_CONFIG_DIR; point it at the shared
# emptyDir (writable by uid 65532 via fsGroup). Without this it tries
# to mkdir /.mc on the read-only-to-nonroot root fs -> "mkdir /.mc:
# permission denied" and every run fails before uploading.
- name: MC_CONFIG_DIR
value: "/work/.mc"
- name: S3_ENDPOINT
valueFrom:
secretKeyRef: { name: forte-drop-secrets, key: S3_ENDPOINT }
- name: S3_BUCKET
value: "drops"
- name: S3_KEY
valueFrom:
secretKeyRef: { name: forte-drop-secrets, key: S3_KEY }
- name: S3_SECRET
valueFrom:
secretKeyRef: { name: forte-drop-secrets, key: S3_SECRET }
volumeMounts:
- name: work
mountPath: /work