I have six nodes setup with keepalived on Ubuntu 18.04. They provide a bunch of VIPs. I can manually failover from one to the other without any issues. Every few days, however, VIPs are lost on all nodes. Keepalived is still running on every node. There are no entries in the logs. I think that it has to do with auto-updates for security related packages. On both of the last to occasions, updates were installed in the morning. It appears that the cause is restarting of systemd-networkd. When this occurs keepalived doesn't get notified that the interface is down. I found this article describing the same symptoms: https://access.redhat.com/solutions/3541681.
To recreate the issue.
1. Start keepalived
2. Verify VIP is present
3. service systemd-networkd restart
4. Verify VIP is missing and did not failover to backup node.
Any suggestions would be greatly appreciated.
Bookmarks