Actions
Bug #58465
closedcephadm: haproxy nfs backends don't properly handle multiple nfs of the same rank
% Done:
0%
Source:
Tags:
backport_processed
Backport:
quincy, pacific
Regression:
No
Severity:
3 - minor
Reviewed:
Description
When the host that an nfs is on goes offline and we deploy a replacement on another host matching the placement. That replacement and the nfs daemon on the offline host will have the same rank. The code to handle getting the IPs for the backend service for the haproxy config doesn't handle this properly and just ends up getting the IP from the node of the nfs of that rank it happens to iterate over last. We should instead always be checking the rank_generated for the nfs daemons and picking the one whose rank_generated is higher.
Actions