Add support for checkpoint images with 'podman run'

[rst0git]

This pull request is a follow up on #15868, #16068 and #16078. It extends the podman run command with support for checkpoint images. When podman run is invoked with an image that contains checkpoint annotation, it would invoke restore.

Example:

podman run -d --name looper busybox /bin/sh -c 'i=0; while true; do echo $i; i=$(expr $i + 1); sleep 1; done'
podman container checkpoint --create-image checkpoint-image-1 looper  
podman run checkpoint-image-1

Does this PR introduce a user-facing change?

The `podman run` command has been extended with support for checkpoint images.

[TomSweeneyRedHat]

quick glance and the changs look reasonable, but the test are very red and angry....

[rst0git]

The following CI failures look unrelated to the changes in this pull request.

Summarizing 4 Failures:

[Fail] Systemd activate [It] stop podman.service 
/var/tmp/go/src/github.com/containers/podman/test/e2e/systemd_activate_test.go:104

[Fail] Podman Info [It] Podman info: check desired network backend 
/var/tmp/go/src/github.com/containers/podman/test/e2e/info_test.go:177

[Fail] Podman Info [It] Podman info: check desired network backend 
/var/tmp/go/src/github.com/containers/podman/test/e2e/info_test.go:177

[Fail] Podman Info [It] Podman info: check desired network backend 
/var/tmp/go/src/github.com/containers/podman/test/e2e/info_test.go:177

Ran 1681 of 1868 Specs in 1258.135 seconds
FAIL! -- 1680 Passed | 1 Failed | 1 Flaked | 0 Pending | 187 Skipped

@edsantiago would you be able to confirm this?

[edsantiago]

f37 root is a network flake, red herring. The common factor, then, is rootless, and it makes no sense. It looks like environment is being clobbered in weird ways. No other PRs are seeing this (i.e., it's likely not something that accident-merged into main), and there's one more weird commonality, a screwup in $CIRRUS_CHANGE_TITLE for all failing runs. I'm going to tentatively blame it on a Cirrus hiccup. I've restarted all failed jobs and will monitor.

[edsantiago]

Nope. My current hypothesis is that the trailing space in the title is getting backslash-escaped improperly in our environment passthrough. @rst0git that will take me some time to debug; ITM, to expedite this, could I ask you to re-push? I've fixed the title (removed trailing space) but it's still necessary to re-push for Cirrus to get the new title. You can git commit --amend, type a character, remove it, and save. That's enough to get a new commit hash for a push. Sorry for the inconvenience.

[rst0git]

Thank you @edsantiago!

[edsantiago]

Submitted PR #16678 to fix the trailing-whitespace bug. Whew! Thanks for pinging me about this, @rst0git.

[rst0git]

@rhatdan @adrianreber PTAL

[openshift-ci]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: rhatdan, rst0git

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [rhatdan]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[adrianreber]

Looks good to me.

[rhatdan]

My one concern with this, is it going to slow down the start of Podman. We are trying to speed up starting of containers, and if this has to inspect images on all containers before starting, this could slow it down.

[adrianreber]

My one concern with this, is it going to slow down the start of Podman. We are trying to speed up starting of containers, and if this has to inspect images on all containers before starting, this could slow it down.

Good point.

@rst0git If I understand the code correctly you are only looking at the annotations of the image and do not actually open the image to detect if it is a checkpoint image, right?

Could you also run a couple of timing tests with and without the tests. How much slower is run with and without your changes.

[rst0git]

I used the following command to measure the performance of podman run, however, the difference in start-up time with and without the changes in this pull request appears to be negligible.

perf stat --repeat 10 podman run alpine:latest true

[rhatdan]

@alexlarsson PTAL
Potential to slow down container startup ,but also chance to speedup application start.

[adrianreber]

I used the following command to measure the performance of podman run, however, the difference in start-up time with and without the changes in this pull request appears to be negligible.

perf stat --repeat 10 podman run alpine:latest true

Can you share your numbers?

[rst0git]

Can you share your numbers?

Yes, I've uploaded the results in https://gist.github.com/rst0git/c3ee40feaa36eb9e5c64750aa92f3b0d

[adrianreber]

The numbers look the same for both cases. So just looking at an annotation as your change does, does not seem to introduce a measurable delay.

Not sure if there is some caching going in Podman when it comes to image annotations. Maybe somebody else can give an additional interpretation for your numbers.

[edsantiago]

This is probably a very stupid comment, but: https://gist.github.com/rst0git/c3ee40feaa36eb9e5c64750aa92f3b0d#file-test-sh-L5

...I would feel so much more comfortable with your results if the string podman were replaced with $PODMAN, and there were two lines immediately above, like

PODMAN=$(pwd)/bin/podman
$PODMAN version

[rst0git]

@edsantiago Thank you for the advice! I've updated the evaluation script:
https://gist.github.com/rst0git/c3ee40feaa36eb9e5c64750aa92f3b0d

[rhatdan]

/lgtm
If this is found to slow things down, we might need to add an option to disable/enable it, but for now I will merge.
Thanks @rst0git