Operating System: All
Platform: All

Hi Luis,

In the above test, do you see any exception in the logs?

(In reply to Shuva Jyoti Kar from comment #1)
> Hi Luis,
>
> In the above test, do you see any exception in the logs?

Also do you connect the switch immediately, or you wait for the leadership election to be complete ?

This bug was exposed after Sanjib's latest cluster test enhancement, you can see it consistently failing in this test:

https://jenkins.opendaylight.org/releng/view/CSIT-3node/job/openflowplugin-csit-3node-clustering-helium-design-only-beryllium/237/robot/report/log.html

• I do not see any specific log messages when the error happens (switch connects and does not get flow).
• We also make sure leadership is being resolved before connecting the switch.
• The issue is only seen with He plugin.

BR/Luis

also is the entity-ownership-shard leadership also finished by this time ? and what if you restart the switch , is the switch getting populated with the switch.

For both questions I modified the test so lets see:

https://git.opendaylight.org/gerrit/#/c/38608/

BR/Luis

From first test results:

https://jenkins.opendaylight.org/releng/job/openflowplugin-csit-verify-3node-clustering-helium-design/69/

• The issue occurs even when all shards (including entity-owner) are resolved.
• It only happens the first time switch connects to old config leader, further connects attempts are good (switch receives flows).

This is consistent with what I have tested locally.

BR/Luis

Last discovery on this bug is that if I add 5 secs between disconnect switch and restart Inv leader, things work stable at least in my setup.

Decreasing priority since no activity for more than 4 months and related to non-default helium design.

No activity on the bug for a while and also He plugin is now deprecated. Please re-open the bug if you see the issue in carbon/nitrogen/oxygen branch.