Hi,

When we made some failure tests with the SNAT, we met a wired issue.

When we quickly generate a start/stop of a vrouter agent where a SNAT VM
was scheduled, that SNAT VM is correctly re-scheduled on another vrouter
agent but the precedent agent continues to announce instance IPs of the
SNAT VM. Both veth of the SNAT VM still exist on the agent but from the
API, the vrouter back refeces of the SNAT VM was correctly updated to the
new vrouter.

If we left few seconds between the stop and the start actions, the issue
does not appear.

Do you have any idea of what could be the issue?

Regards,
Édouard.
_______________________________________________
Dev mailing list
[email protected]
http://lists.opencontrail.org/mailman/listinfo/dev_lists.opencontrail.org

Reply via email to