remove multipathd socket workaround

[miabbott]

The fixed systemd unit for the multipathd.service landed in RHEL
8.4.z as part of device-mapper-multipath-0.8.4-10.el8_4.3 and it is
landing in RHEL 8.4.z, which will be consumed by RHCOS 4.10.

See: https://bugzilla.redhat.com/show_bug.cgi?id=2054876

[miabbott]

We can't make this change to testing-devel yet, as the fixed version of device-mapper-multipath isn't scheduled until F36

https://bugzilla.redhat.com/show_bug.cgi?id=2008098

[miabbott]

I sanity tested this with an RHCOS 4.10 build locally and ran ext.config.shared.multipath.multipathd-socket-fix

$ kola run ext.config.shared.multipath.multipathd-socket-fix
⚠️  Skipping kola test pattern "coreos.ignition.journald-log":
  👉 https://github.com/coreos/coreos-assembler/issues/1173
=== RUN   non-exclusive-tests
=== RUN   non-exclusive-tests/ext.config.shared.multipath.multipathd-socket-fix
--- PASS: non-exclusive-tests (26.53s)
    --- PASS: non-exclusive-tests/ext.config.shared.multipath.multipathd-socket-fix (1.39s)
PASS, output in tmp/kola/qemu-unpriv-2022-04-22-1551-39

[miabbott]

Closing for #1685

[dustymabe]

We can't make this change to testing-devel yet, as the fixed version of device-mapper-multipath isn't scheduled until F36

https://bugzilla.redhat.com/show_bug.cgi?id=2008098

I don't see why we can't just adjust the current logic in testing-devel to only apply to RHEL < 8.4 or just exit 0 if RHEL at all (assuming all RHCOS have move past 8.4). Then we can just cherry-pick that change to each rhcos branch we need it to apply to.

[miabbott]

Just kidding

[miabbott]

I don't see why we can't just adjust the current logic in testing-devel to only apply to RHEL < 8.4 or just exit 0 if RHEL at all (assuming all RHCOS have move past 8.4).

Sure we can try that for testing-devel, but the current state of the 4.10 builds are blocking us from releasing RHCOS 4.9/4.8/4.7, so dropping the workaround completely here should be fine.

[dustymabe]

LGTM

[miabbott]

Looks like we need more picks

[coreos-assembler]$ kola run
Error: Parsing metadata from src/config/tests/kola/shared/root-reprovision/swap-before-root/test.sh: parsing # kola: { "distros": "fcos", "platforms": "qemu", "minMemory": 4096, "timeoutMin": 15, "allowConfigWarnings": true }
: json: unknown field "allowConfigWarnings"
2022-04-22T18:51:37Z cli: Parsing metadata from src/config/tests/kola/shared/root-reprovision/swap-before-root/test.sh: parsing # kola: { "distros": "fcos", "platforms": "qemu", "minMemory": 4096, "timeoutMin": 15, "allowConfigWarnings": true }
: json: unknown field "allowConfigWarnings"

[miabbott]

Or rather, we need some picks for cosa:rhcos-4.10

[dustymabe]

hmm OK. Maybe I was mistaken.. Here's the first error we saw:

[2022-04-22T08:50:21.997Z] kola -p qemu-unpriv --output-dir /srv/fcos/tmp/fcosKola-EaHcB/kola run --rerun --qemu-nvme=true basic
[2022-04-22T08:50:22.251Z] 2022-04-22T08:50:22Z platform/conf: warnings parsing config: warning at $.storage.disks.0.partitions.1.label, line 13 col 18: incorrect partition number; a new partition will be created using reserved label
[2022-04-22T08:50:22.251Z] Error: Parsing config.ign: configured to treate config warnings as fatal

I assumed it was because COSA had been updated but the tests hadn't and thought we needed 60c082c

That's clearly not the case (you can drop that cherry-picked commit).

This ignition warning is coming from the basic test, which is odd.

[miabbott]

We never properly enabled the CI on these rhcos-4.x branches...since nothing is going to build FCOS from these branches, we should probably just stop running these tests entirely

[miabbott]

I'll run the full kola run suite to confirm things will be OK

[miabbott]

OK, full kola run was successful

[miabbott]

xref: openshift/os#786