Dear Maintainer,

I'm currently working on updating the etcd-server package from 3.5.16 -> 3.5.22, and
have uploaded the result to debian/experimental. The new package brings a number of
upstream bugfixes, and I'd like to upload it to sid.


While testing reverse dependencies, I've noticed that the autopkgtest in the patroni
package fail when building against that new version of etcd:

1269s autopkgtest [12:08:51]: @@@@@@@@@@@@@@@@@@@@ summary
1269s test                 PASS
1269s acceptance-etcd3-basic FAIL non-zero exit status 1
1269s acceptance-etcd3-full FAIL non-zero exit status 1
1269s acceptance-etcd-basic PASS
1269s acceptance-zookeeper-basic PASS
1269s acceptance-raft-basic PASS


This seems to happen on all architectures, here are the logs for amd64 and i386:

https://ci.debian.net/packages/p/patroni/unstable/i386//63284595/
https://ci.debian.net/packages/p/patroni/unstable/amd64//63284593/


I'm having a really hard time reading the output of the tests. It seems that
autopkgtest is executing behavioral tests using the python "behave" package, which
I'm unfamiliar with. Please assist with diagnosing this issue and let me know
if that is an issue with the test or a regression in the etcd package.


BTW, I've noticed this failure using debusine, cf
https://debusine.debian.net/debian/developers/work-request/132304/ it
seems that all other reverse depends pass, with the exception of
vip-manager2, which I'm going to look at next.

Hi,

Thanks for that and thanks for this report!

Looks like it is this error:

|File "/usr/lib/python3/dist-packages/patroni/dcs/etcd3.py", line 461, in set
|     ret = not old_value or int(old_value['mod_revision']) < int(value['mod_revision'])
|                                ~~~~~~~~~^^^^^^^^^^^^^^^^
|KeyError: 'mod_revision'
[...]
|patroni.dcs.etcd3.Etcd3Error: unexpected error
|2025-08-16 11:51:12,108 INFO [/usr/lib/python3/dist-packages/patroni/__main__.py:215 - _run_cycle]: Unexpected exception raised, please report it as a BUG

I checked with Patroni upstream about this and they said they'd tried
with both 3.5.22 and latest 3.6 and see no problem, so this does not
seem to be a general error, I'll try to reproduce this and look further
into it.


Michael

Hi,

Maybe it's a toolchain issue, did you try rebuilding both the unstable
and experimental version on either testing or unstable and see whether
that changes anything?

Also, I guess that means the etcd package building process should run
more tests...


Michael