[stable][ovn] frequent OVN DB leader changes increase rate of Neutron API errors
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
neutron |
Fix Released
|
Medium
|
Frode Nordahl |
Bug Description
Open vSwitch v2.16 introduces a change of behavior for clustered databases,
before the leader creates a snapshot of the database it will transfer
leadership to a different server.
For versions of Neutron where the Southbound DB IDL requires a connection to
the leader this will cause frequent reconnection. In a loaded system this will
subsequently lead to increased rate of API errors for the neutron-server
and delays for the neutron-
instances not getting metadata.
In the main branch Neutron has recently changed to not require a connection to
the leader for the Southbound DB [0].
End users can choose to use more recent versions of OVS and OVN than what was
available at the time of release of past OpenStack releases, so we would like
to have this change backported to Ussuri.
This bug is to track the backport of the change.
tags: | added: ovn |
Changed in neutron: | |
status: | New → Confirmed |
importance: | Undecided → Medium |
assignee: | nobody → Frode Nordahl (fnordahl) |
Fix proposed to branch: stable/xena /review. opendev. org/c/openstack /neutron/ +/829486
Review: https:/
Fix proposed to branch: stable/wallaby /review. opendev. org/c/openstack /neutron/ +/829487
Review: https:/
Fix proposed to branch: stable/victoria /review. opendev. org/c/openstack /neutron/ +/829488
Review: https:/
Fix proposed to branch: stable/ussuri /review. opendev. org/c/openstack /neutron/ +/829489
Review: https:/