Bug#1017527: acceptance-* tests should be marked as 'flaky'

2022-08-17 Thread Christoph Berg
Re: Pierre Gruet
> The three autopkgtests acceptance-consul, acceptance-etcd and
> acceptance-zookeeper began failing at a significant frequency recently, at
> least on armhf, independently of each other.

I wonder if we shouldn't strip down these tests to some basic smoke
test, for example by running the first "basic replication" test only.

While that of course reduces the test surface, it will hopefully get
rid of the test failures we've been seeing. (Did the tests actually
ever find a real bug as opposed to just the tests failing
intermittently?)

> Please see the page
> https://ci.debian.net/packages/p/patroni/testing/armhf/
> and the bottoms of the various test logs for an illustration.

... and https://pgdgbuild.dus.dg-i.net/view/Binaries/job/patroni-binaries/

Christoph



Bug#1017527: acceptance-* tests should be marked as 'flaky'

2022-08-17 Thread Pierre Gruet
Source: patroni
Version: 2.1.4-1
Severity: normal
Control: affects -1 src:zookeeper src:psycopg2

Dear Maintainer,

The three autopkgtests acceptance-consul, acceptance-etcd and
acceptance-zookeeper began failing at a significant frequency recently, at
least on armhf, independently of each other. This makes the migration of
reverse-dependencies to testing rather difficult, as one basically has to retry
the CI tests again and again until the three above-quoted tests pass.

Please see the page
https://ci.debian.net/packages/p/patroni/testing/armhf/
and the bottoms of the various test logs for an illustration.

I think the three acceptance-* tests should be marked with
Restrictions: flaky
so that they allow for smooth testing migrations of the reverse dependencies.

Thanks a lot for considering,

Best regards,

-- 
Pierre