kuryr-kubernetes

  1. Bug #1778516
  2. Comment #2

Comment 2 for bug 1778516

Revision history for this message
Yossi Boaron (yossi-boaron-1234) wrote : Re: Amphora fails to start with IP address already allocated error

Seems that the root cause for this specific case was kuryr-controller's pod being restarted periodically (cause of health check failure).
So, this due to this restarts :

1. services/LB created by Kuryr-controller weren't annotated properly.
2. When Kuryr restarted, since K8S service's annotation wasn't saved properly, LBaaS handler called LBaaS driver 'ensure_loadbalancer' function to re-create LB, and this function doesn't handle properly this case.
In most of the cases, LBaaS handler won't call LBaaS driver in that case.

So, in the bottom line :

A. The bug description should change to
LBaaS driver doesn't handler properly 'ensure_loadbalancer' for existing LB.

B. I don't think that bug's severity should be critical.