Description of problem: Currently nbdb/sbdb election-timer value is set to 5 second (from default 1 second) during the pod startup. After pods are in running state, and user manually changes the value of election-timer it's stored in the db. But when the pod restarts, it actually override the election-timer to default 5 second again, ideally it should not reset the value if the election-timer value is manually overridden. Version-Release number of selected component (if applicable): How reproducible: Very easy to reproduce. Steps to Reproduce: *.* Deploy ovn-kubernetes *.* Change the nbdb/sbdb election-timer to some value except default value of 5 second (e.g 10 seconds) *.* Restart the nbdb/sbdb pod. It will reset the value from 10 second to 5 second. Following log message will be logged in the nb/sb db log file ```2020-07-10T14:58:29Z|00025|raft|INFO|Election timer changed from 16000 to 5000``` Actual results: Restart resets the election-timer value to default value. Expected results: It should restart the pods with the non-default value (if configured) Additional info:
There is a reverted PR on this https://github.com/openshift/cluster-network-operator/pull/769 which presumably causing https://bugzilla.redhat.com/show_bug.cgi?id=1872098. There might be other reasons which are still being investigated behing OVN cluster failures. Moving this to assigned for now, awaiting resolution on PR
We do not support people editing the NBDB directly. Don't shoot yourself in the foot.