Support auto updating containers running inside pods. Similar to
containers, the systemd units need to be generated via
`podman-generate-systemd --new $POD` to generate the pod's units.
Note that auto updating a container inside a pod will restart the entire
pod. Updates of multiple containers inside a pod are batched, such that
a pod is restarted at most once. That is effectively the same mechanism
for auto updating containers in a K8s YAML via the `podman-kube@`
template or via Quadlet.
Updating a single container unit without restarting the entire pod is
not possible. The reasoning behind is that pods are created with
--exit-policy=stop which will render the pod to be stopped when auto
updating the only container inside the pod. The (reverse) dependencies
between the pod and its containers unit have been carefully selected for
robustness. Changes may entail undesired side effects or backward
incompatibilities that I am not comfortable with.
Fixes: #17181
Signed-off-by: Valentin Rothberg <vrothberg@redhat.com>
Very belated successor to #14046.
I don't know why this is so important to me. Probably because we're
doing a halfhearted sloppy job of documenting, and new options get
added, and not documented, and that's just wrong.
I've given up on documenting internal structs. This iteration
has a $Format_Exceptions table defined at the top of the xref
script, enumerating a hardcoded defined set of podman commands
and fields that should remain undocumented.
This iteration also forgives completely-undocumented formats.
If podman-foo has a --format, but podman-foo.1.md does not
list *any* valid fields, the script warns but does not fail.
This at least is better than documenting a random mix of fields.
This version of the xref script is much slower: 10s vs 4. I
think we can live with that in a CI-only script.
Signed-off-by: Ed Santiago <santiago@redhat.com>
Add auto-update support to `podman kube play`. Auto-update policies can
be configured for:
* the entire pod via the `io.containers.autoupdate` annotation
* a specific container via the `io.containers.autoupdate/$name` annotation
To make use of rollbacks, the `io.containers.sdnotify` policy should be
set to `container` such that the workload running _inside_ the container
can send the READY message via the NOTIFY_SOCKET once ready. For
further details on auto updates and rollbacks, please refer to the
specific article [1].
Since auto updates and rollbacks bases on Podman's systemd integration,
the k8s YAML must be executed in the `podman-kube@` systemd template.
For further details on how to run k8s YAML in systemd via Podman, please
refer to the specific article [2].
An examplary k8s YAML may look as follows:
```YAML
apiVersion: v1
kind: Pod
metadata:
annotations:
io.containers.autoupdate: "local"
io.containers.autoupdate/b: "registry"
labels:
app: test
name: test_pod
spec:
containers:
- command:
- top
image: alpine
name: a
- command:
- top
image: alpine
name: b
```
[1] https://www.redhat.com/sysadmin/podman-auto-updates-rollbacks
[2] https://www.redhat.com/sysadmin/kubernetes-workloads-podman-systemd
Signed-off-by: Valentin Rothberg <vrothberg@redhat.com>
Refactor the --authfile option.
My suggestion for review:
1) run hack/markdown-preprocess-review and immediately Ctrl-Q to
quit out of diffuse, which is completely unusable for this
many files; then
2) cd /tmp/markdown-preprocess-review.diffs/authfile
- this is the directory created by the review script
3) rm podman-image-sign* podman-log* podman-search.1.md.in
- because they're essentially identical to podman-create
4) rm podman-manifest-* podman-push.*
- because they're 100% identical to podman-kube-play
5) rm podman-kube-play*
- because it's apart-from-whitespace identical to podman-build
(use "wdiff" to confirm)
6) rm podman-auto-update*
- because that's the one I chose (hence == zzz-chosen.md)
(You should obviously run your own diff/cmp before rm, to confirm
my assertions about which files are identical).
After all that, you have a manageable number of files which
you can scan, read, diff against zzz-chosen.md, even run diffuse.
This option is IMHO the poster child for why we need this kind
of man page refactoring.
Signed-off-by: Ed Santiago <santiago@redhat.com>
Valentin Rothberg
Ed Santiago
Daniel Lublin
OpenShift Merge Robot
Andrew Denton