[sys] kube play read-only: racy status checks

[edsantiago]

Flake seen in my hammer-sqlite PR, f37 aarch64 root, both local and remote:

not ok 535 podman kube play read-only from containers.conf
...
# podman inspect --format {{.State.ExitCode}} pod1-test2
0
#/vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv
#|     FAIL: Container / should be read/only   <---- poor message. Should read "exit status from touch /file on RO container"
#| expected: '1'
#|   actual: '0'

Root cause: when podman kube play runs containers, a container that exits nonzero seems to be infinite-looped: it keeps retrying forever. The above inspect caught it in a moment when the container had just started.

Solution: I give up. I've spent an hour trying to play with restartPolicy and container create --restart=no (which does not get seen by podman generate kube) and looking at man pages. I can find no way to tell podman-kube to just run that container once.

I'm astonished that this doesn't flake more often.

[rhatdan]

@umohnani8 PTAL

[edsantiago]

One of our most popular flakes this week. I'm hoping #17627 will let me find a way to fix this.

• debian-12 : sys podman debian-12 root host boltdb
  • PR CI: enable sqlite system tests #17954
    • 04-06 17:04 in [sys] podman kube play read-only from containers.conf
• debian-12 : sys remote debian-12 root host boltdb [remote]
  • PR CI: enable sqlite system tests #17954
    • 04-07 15:06 in [sys] podman kube play read-only
• fedora-37-aarch64 : sys podman fedora-37-aarch64 root host boltdb
  • PR build(deps): bump github.com/spf13/cobra from 1.6.1 to 1.7.0 #18061
    • 04-06 08:39 in [sys] podman kube play read-only
• fedora-37-aarch64 : sys podman fedora-37-aarch64 root host sqlite
  • PR Ed's pet PR with no flake retries #17831
    • 04-07 16:06 in [sys] podman kube play read-only from containers.conf
    • 04-06 16:57 in [sys] podman kube play read-only
    • 04-04 12:10 in [sys] podman kube play read-only from containers.conf
• fedora-37-aarch64 : sys remote fedora-37-aarch64 root host boltdb [remote]
  • PR fix: Document removing anonymous volumes at create #18151
    • 04-11 21:32 in [sys] podman kube play read-only from containers.conf
  • PR api: auth: fix nil deref #18143
    • 04-12 06:15 in [sys] podman kube play read-only from containers.conf
  • PR Read kube_generate_type from containers.conf #18140
    • 04-11 10:02 in [sys] podman kube play read-only from containers.conf
  • PR Fix up codespell errors #18105
    • 04-08 01:39 in [sys] podman kube play read-only from containers.conf
  • PR The --ulimit option now also accepts the name with an RLIMIT_ prefix both upper and lower case. #18078
    • 04-06 17:26 in [sys] podman kube play read-only from containers.conf
    • 04-06 11:07 in [sys] podman kube play read-only
  • PR chore: replace github.com/ghodss/yaml with sigs.k8s.io/yaml #18066
    • 04-05 12:01 in [sys] podman kube play read-only
  • PR CI: enable sqlite system tests #17954
    • 04-12 07:44 in [sys] podman kube play read-only from containers.conf
    • 04-07 15:10 in [sys] podman kube play read-only from containers.conf
• fedora-37-aarch64 : sys remote fedora-37-aarch64 root host sqlite [remote]
  • PR Ed's pet PR with no flake retries #17831
    • 04-05 10:43 in [sys] podman kube play read-only from containers.conf
    • 04-04 11:59 in [sys] podman kube play read-only