let me know if I can do anything to test/debug that would help
skitt said during the Kernel call that this must be another case of those pesky feature reload problems - the SSH server in Karaf restarts as an (unintended) consequence of the feature:install odl-netvirt-openstack ... skitt usually can figure out what's causing those bundle reloads and conjure up a fix. In the mean time though dfarrell07 and jluhrsen, I bet you guys that using feature:install --no-auto-refresh odl-netvirt-openstack probably works around this problem?
> I bet you guys that using feature:install --no-auto-refresh odl-netvirt-openstack probably works around this problem?
vorburger I tried it locally and in the sandbox, and sadly no it didn't work around it.
https://jenkins.opendaylight.org/sandbox/job/packaging-test-rpm-master/22/console
Here's another karaf.log, but this time I waited longer before collecting it (since ODL was still running after it dropped my SSH connection)
It now shows some warnings related to blueprint, maybe relevant?
I guess this should be a blocker for Oxygen, right? It's a major, user-facing regression from Nitrogen. klou
jluhrsen et al - can you just sanity check that you also see this locally? It should be easy to reproduce.
yeah, I can reproduce this:
- sudo systemctl start opendaylight
- ssh -p 8101 karaf@127.0.0.1
- feature:install odl-openstack-netvirt
(after aprox 60s, the ssh connection is closed)
I can ssh back in though, so maybe this is not a complete blocker. But it does break our automation
and tests so it would be nice to get fixed.
dfarrell07, maybe we can change our tests to re-connect and continue it's work, instead of failing
when the ssh connection is dropped.
I also confirmed that his problem is not in a recent Nitrogen distro.
I have marked this issue as a blocker for Oxygen.
dfarrell07 yes I can reproduce it too — that’s why I marked it “confirmed” a few days ago 
.
Please provide an update on this issue. Is this a true blocker? Any ETA on resolution?
Talked with skitt about this and we seem to agree that it should be downgraded from blocker. It's a major bug that may impact downstream automation, but it's not going to be easy/realistic to fix quickly, and since there's a work-around (reconnect to SSH) maybe it shouldn't block the release.
I cannot reproduce this with odlparent-5.0.0