examples/demo: fix two real bugs this exact demo just hit live
CI / chart (push) Successful in 1s
CI / security (push) Successful in 3m24s
CI / test (push) Successful in 10m45s

1. Renamed every object this demo creates (TerdutServer, Postgres
   Secret/Deployment/Service) from terdut-demo[-postgres] to
   terdut-operator-demo[-postgres]. The user applied this kit into the
   already-live "terdut-demo" namespace -- the real operator exercise
   from earlier in this repo's own history -- and this demo's own
   TerdutServer/Postgres objects shared that exact name. The TerdutServer
   apply was rejected outright (DatabaseSpec's own CEL rule: adding dsn
   while the live object already had postgresClusterRef violates "exactly
   one of" and the API server refused it), and the real Postgres Service
   was never touched (confirmed live: still Zalando's own spilo selector,
   endpoint still the real StatefulSet pod) -- but the Postgres Secret and
   Deployment, having no such protection, were created as brand new,
   extra, crash-looping objects sitting right next to the real ones.
   Prefixing every name this demo creates means a repeat of this exact
   mistake no longer collides with anything, documented directly in
   README.md now.

2. The actual crash itself, independent of (1): capabilities.drop: ["ALL"]
   (added responding to a PodSecurity "restricted" warning) took
   CAP_CHOWN/CAP_FOWNER away from the root user postgres:17-alpine's own
   entrypoint needs to chown/chmod the data directory before it drops
   privileges itself -- confirmed in a real crashed pod's logs: `chmod:
   /var/run/postgresql: Operation not permitted`. kubectl apply
   --dry-run=server, which is as far as this got verified before, only
   checks admission policy; it was never actually booted. Removed the
   capability drop and verified for real this time: applied just
   00-postgres.yaml alone into a disposable namespace, waited for the pod
   to go Ready, read its logs ("database system is ready to accept
   connections"), then deleted that namespace.
This commit is contained in:
Niklas Ye
2026-10-02 13:25:47 +02:00
parent 5b45cf72e1
commit d9315322fc
6 changed files with 44 additions and 28 deletions
+21 -16
View File
@@ -10,7 +10,7 @@
apiVersion: v1
kind: Secret
metadata:
name: terdut-demo-postgres
name: terdut-operator-demo-postgres
type: Opaque
stringData:
password: demo-not-a-real-password
@@ -18,9 +18,9 @@ stringData:
apiVersion: apps/v1
kind: Deployment
metadata:
name: terdut-demo-postgres
name: terdut-operator-demo-postgres
labels:
app: terdut-demo-postgres
app: terdut-operator-demo-postgres
spec:
replicas: 1
# Recreate, not RollingUpdate: emptyDir means a new pod starts with an
@@ -30,27 +30,32 @@ spec:
type: Recreate
selector:
matchLabels:
app: terdut-demo-postgres
app: terdut-operator-demo-postgres
template:
metadata:
labels:
app: terdut-demo-postgres
app: terdut-operator-demo-postgres
spec:
containers:
- name: postgres
image: postgres:17-alpine
# Partial, deliberately: the official image's entrypoint needs to
# start as root to chown the data directory before it drops
# start as root to chown/chmod the data directory before it drops
# privileges itself (gosu, to the postgres user) -- forcing
# runAsNonRoot here would just refuse to start the container. A
# "restricted" PodSecurity namespace warns on that gap rather
# than blocking (confirmed server-side against this operator's
# own dev cluster), which is an acceptable tradeoff for Postgres
# that exists only to be thrown away with the rest of this demo.
# runAsNonRoot would just refuse to start the container, and
# dropping all capabilities (an earlier version of this file did)
# takes CAP_CHOWN/CAP_FOWNER away from that same root user, which
# is a different way of breaking the identical startup step:
# confirmed the hard way, as `chmod: /var/run/postgresql:
# Operation not permitted` in a real pod's logs, not caught by
# `kubectl apply --dry-run=server` -- that only checks admission
# policy, never whether the container actually boots. A
# "restricted" PodSecurity namespace warns on the remaining gap
# (no runAsNonRoot) rather than blocking, which is an acceptable
# tradeoff for Postgres that exists only to be thrown away with
# the rest of this demo.
securityContext:
allowPrivilegeEscalation: false
capabilities:
drop: ["ALL"]
seccompProfile:
type: RuntimeDefault
ports:
@@ -64,7 +69,7 @@ spec:
- name: POSTGRES_PASSWORD
valueFrom:
secretKeyRef:
name: terdut-demo-postgres
name: terdut-operator-demo-postgres
key: password
volumeMounts:
- name: data
@@ -81,10 +86,10 @@ spec:
apiVersion: v1
kind: Service
metadata:
name: terdut-demo-postgres
name: terdut-operator-demo-postgres
spec:
selector:
app: terdut-demo-postgres
app: terdut-operator-demo-postgres
ports:
- name: postgres
port: 5432