-> Optional (temp) Workaround for Kubernetes + CloudNativePG (CNPG)
users

This regression prevents container processes from sending SIGURG under
the forced stacked profile cri-containerd.apparmor.d//&unconfined.
PostgreSQL uses SIGURG for latch wakeups, so hot standbys hang in
recovery (clients get "the database system is starting up", the startup
process sits in epoll_wait, no walreceiver) and the primary can be
affected too.

You don't need to disable AppArmor for the node or for other workloads.
You can make only the CNPG database pods run unconfined:

```
spec:
  # ... existing spec ...

  # Workaround for LP #2172640: SIGURG denied under the stacked
  # cri-containerd.apparmor.d//&unconfined profile.
  # AppArmor is Unconfined for the postgres/instance-manager container only;
  # every other workload keeps the default profile.
  #
  # NOTE: CNPG's spec.securityContext REPLACES the operator's hardened
  # container defaults, so the defaults are repeated here on purpose.
  #
  # TODO: remove this whole block once the nodes run a kernel with the
  # LP #2172640 fix (see the bug for the fixed version) and/or containerd
  # includes containerd/containerd#14487.
  securityContext:
    appArmorProfile:
      type: Unconfined
    allowPrivilegeEscalation: false
    capabilities:
      drop:
        - ALL
    privileged: false
    readOnlyRootFilesystem: true
    runAsNonRoot: true
    seccompProfile:
      type: RuntimeDefault
```

AppArmor is applied at container creation, so existing pods keep the old
profile. The pods must be recreated: CNPG's rolling update normally does
this (replicas first, primary last; primaryUpdateStrategy: unsupervised
is the default). If the operator doesn't proceed, delete the replicas'
pods one at a time, waiting for 2/2 each, then the primary last.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2172640

Title:
  Regression in 7.0.0-42: AppArmor forced "//&unconfined" stack on
  container processes blocks signals between them (PostgreSQL hangs)

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2172640/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to