Revert "Make sure that after toggling ssl on/off all services are restarted #3365
Conversation
…tarted" This reverts commit 9d523e7.
|
no the problem is not the maintenance mode but the problem is neutron-ha-tool crashing due to not being restarted. we can fix that by deleting the failcounter as I did previously, but that was rejected in the previous review: #3286 (review) |
|
I am working on a patch to fix the rootcause, if you don't have time for that then we can use the crm_failcount removal hack as an interim solution |
|
crowbar/crowbar-openstack#2112 is the fix wip for this |
|
So why did this only start to fail after #3286 got merged (that's the reason I opened this PR - to make sure it really isn't the culprit)? Over the weekend before April 15th there were a whole bunch of successful HA job runs, and for |
|
@jgrassler very simple, because neutron-ha-tool needs to crash repeatedly several times in a row (like 15 times in a row or so) before pacemaker logs a failure. and in the previous code the crash just happened later, interrupting tempest and other things, than clearly causing a repeatable single fail. now it waits, and you trigger it all the time. before it was only happening when you were particularly unlucky. |
|
also, in most cases the "failcount" check was already executed before the failure was logged (and it does so only once) |
There was a problem hiding this comment.
please spend your review electrons here: crowbar/crowbar-openstack#2112
This reverts commit 9d523e7.
Reverting this should fix HA job breakage.