flaky test: TestCheckpoint(actuated-arm64-6cpu-8gb, 1.21.x)

[lifubang]

Sometimes fail:

=== RUN   TestCheckpoint
    checkpoint_test.go:115: === /tmp/TestCheckpoint1387080808/003/criu-parent/dump.log ===
    checkpoint_test.go:115: open /tmp/TestCheckpoint1387080808/003/criu-parent/dump.log: no such file or directory
    checkpoint_test.go:119: CRIU feature check failed
--- FAIL: TestCheckpoint (0.24s)

Sometimes success to skip:

=== RUN   TestCheckpoint
    checkpoint_test.go:115: === /tmp/TestCheckpoint194762558/003/criu-parent/dump.log ===
    checkpoint_test.go:115: open /tmp/TestCheckpoint194762558/003/criu-parent/dump.log: no such file or directory
    checkpoint_test.go:117: criu is missing features: [MemTrack]
--- SKIP: TestCheckpoint (0.51s)

[kolyshkin]

I am working on that. The issue is, criu feature check is not doing what it should.

[kolyshkin]

Interesting; I've never seen it being able to skip! I guess it was skipping when using a runner with an older kernel.

I found two reasons for the test to fail.

1. actuated-arm64 currently runs a kernel with CONFIG_NF_TABLES_INET not set, which uncovers a bug in criu (see criu: fix a fatal failure if nft doesn't work checkpoint-restore/criu#2403).

2. It seems that the feature check test is not working like it should. I've added a patch to not test pre-dump if memory tracking is not available (one of commits in PR fix checkpoint/restore tests on actuated-arm64 #4276).

These two PRs address the issue.

[alexellis]

@kolyshkin - related question for you and @AkihiroSuda as part of the runc project - checkpoint-restore/criu#2402 (comment)