Vendor c/image after https://github.com/containers/image/pull/1816

[mtrmac]

WIP, this is primarily to expose the code to flakes, and to collect production-environment error values and other data.

Look for BODYREADER DEBUG messages in logs, they are intentionally very loud right now

Does this PR introduce a user-facing change?

Automatically transparently resumes aborted pull connections (in some strictly limited situations)

[mtrmac]

Cc: @edsantiago

[mtrmac]

@edsantiago the test logs (very reasonably) don’t include output from Podman commands ran as part of the tests if the tests succeed. Do you know if there is some pre-existing, reasonably convenient, trick I could use to log things from deep inside Podman code in a way that would then be possible to view on test success? Maybe syslog?

[mtrmac]

dial tcp: lookup cdn03.quay.io: no such host are quite frequent, and not the target of this PR.

I mean, are we supposed to detect and workaround consistently incorrect DNS in redirect targets? It’s probably possible (after we successfully ping quay.io/v2/, any future “DNS not found” errors can be assumed to be infrastructure), but … seriously?

[edsantiago]

some pre-existing, reasonably convenient, trick I could use to log things from deep inside Podman code in a way that would then be possible to view on test success? Maybe syslog?

From a BATS script, you can echo to fd3, prepending a hash:

    echo "# debug: $output" >&3

From podman itself, you can't, because I take pains to close fd3 before running podman. But journal logs are available in CI, via a series of clicks on the Cirrus page (near the bottom of the accordions).

I mean, are we supposed to detect and workaround consistently incorrect DNS in redirect targets?

I reported the flake to quay on Thursday. Haven't made it past the "works for me" phase yet.

[mtrmac]

One caught failure:

BODYREADER DEBUG: Reading blob body, decision inputs: lastRetryOffset 0, offset 60631823, error &errors.errorString{s:\"unexpected EOF\"}

Good, so the heuristics did match at least in this one case.

[mtrmac]

BODYREADER DEBUG: Reading blob body, decision inputs: lastRetryOffset 0, offset 260352, error &errors.errorString{s:"unexpected EOF"}

The current heuristic requires 1 MB, and did not trigger here.

[mtrmac]

Analyzing test runs from Jan 30 (and making a record here so that we have some sort of statistics):

Quite a few tests are broken by the extra logging; I need to tune that down.

---

BODYREADER DEBUG: Reading blob body, decision inputs: lastRetryOffset 0, offset 2771195, time diff 0 ms, error &errors.errorString{s:\"unexpected EOF\"}

so “immediately” after successfully receiving data (possibly buffered locally? OTOH clearly not a stalled connection), we get unexpected EOF.

On the positive side, reconnecting seems to have worked (the first? recorded instance).

---

BODYREADER DEBUG: Reading blob body, decision inputs: lastRetryOffset 0, offset 7403776, time diff 0 ms, error &errors.errorString{s:\"unexpected EOF\"}"

also seems to have worked.

[mtrmac]

vendor/github.com/openshift/imagebuilder/dockerfile/parser/parser.go:122:12: undefined: system.LCOWSupported

openshift/imagebuilder#238

Note to self: a backport to the 5.24 branch might be in order…

[mtrmac]

Backported, let’s see…

(This PR is currently primarily helpful as a source of frequent enough failures to gather more data. Actually backporting depends on following a process…)

[vrothberg]

LGTM

I am also in favor of a backport. If we backport, today is a good time to get it in before the 4.4.1 release.

[mtrmac]

Merging #17342 makes this PR, going back to a c/image 5.24 stable branch, plausible but less unattractive. Something like the closed #17405, rolling c/image even further forward, would make Podman smaller (and on the main branch it’s closer to the long-term direction).

[mtrmac]

Due to #17342 being merged, just using a c/image 5.24 branch would be a revert. So, this PR now just rolls c/image forward to include the main-branch merged PR. That brings several other dependency updates.

So, this also:

• includes unreleased system: use containers/storage for LCOW validation openshift/imagebuilder#246 to work with the updated docker/docker dependency.
• updates some references to newly deprecated docker/docker symbols.

Ready for review and possible merging.

[edsantiago]

podman grew by 91528 bytes; max allowed is 51200

Ouch.

Restarted system-test flake, it's the dial tcp: lookup cdn03.quay.io: no such host one.

[vrothberg]

LGTM
@edsantiago PTAL

[openshift-ci]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: mtrmac, vrothberg

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 [mtrmac,vrothberg]

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

[edsantiago]

I'm sorry... this is not something I can realistically review, given my ignorance of the underlying modules.

[vrothberg]

@rhatdan PTAL

[vrothberg]

@containers/podman-maintainers PTAL

[rhatdan]

/lgtm