commit 3c86ec443f9098b4a41c03ef6295d6af376d35a5
Author: Eric MacDonald <email address hidden>
Date: Thu Jun 6 21:12:54 2019 -0400
Make heartbeat cluster state change notifications immediate
Investigation of the reported issue lead to the discovery
that the heartbeat service was not sending a cluster state
change notification after detecting that the peer controller
was no longer providing its cluster state view.
This update makes the following two cluster state change
notification improvements to ensure that SM is notified
immeditely following all cluster state changes.
1. send SM current cluster state view notification as soon
as the state change is detected rather than waiting for
the end of the heartbeat period.
2. consider the loss of the peer controller cluster view as
a cluster state change that, like other state changes,
warrants an immeduiate notification.
Test Plan:
PASS: Inspect heartbeat service logs for expected behavior.
PASS: Reboot the active controller 10+ times and verify SM
switches activity to the inservice peer controller in
a timely manner.
Change-Id: Iebf466a6d75e2f56fae1fdbdfc148127db139c99
Closes-Bug: 1830774
Signed-off-by: Eric MacDonald <email address hidden>
Reviewed: https:/ /review. opendev. org/663813 /git.openstack. org/cgit/ starlingx/ metal/commit/ ?id=3c86ec443f9 098b4a41c03ef62 95d6af376d35a5
Committed: https:/
Submitter: Zuul
Branch: master
commit 3c86ec443f9098b 4a41c03ef6295d6 af376d35a5
Author: Eric MacDonald <email address hidden>
Date: Thu Jun 6 21:12:54 2019 -0400
Make heartbeat cluster state change notifications immediate
Investigation of the reported issue lead to the discovery
that the heartbeat service was not sending a cluster state
change notification after detecting that the peer controller
was no longer providing its cluster state view.
This update makes the following two cluster state change
notification improvements to ensure that SM is notified
immeditely following all cluster state changes.
1. send SM current cluster state view notification as soon
as the state change is detected rather than waiting for
the end of the heartbeat period.
2. consider the loss of the peer controller cluster view as
a cluster state change that, like other state changes,
warrants an immeduiate notification.
Test Plan:
PASS: Inspect heartbeat service logs for expected behavior.
PASS: Reboot the active controller 10+ times and verify SM
switches activity to the inservice peer controller in
a timely manner.
Change-Id: Iebf466a6d75e2f 56fae1fdbdfc148 127db139c99
Closes-Bug: 1830774
Signed-off-by: Eric MacDonald <email address hidden>