ARP driver in astra-appliance needs additional exception handling
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Astara |
Confirmed
|
Medium
|
Ryan Petrello |
Bug Description
The appliance has code that, at configuration time, looks at each network and removes ARP entries that no longer have networks associated with them. This can happen, for example, if a network is hot-detached from the backing Nova VM and the router still has a stale entry:
https:/
This code needs some exception handling around the `/usr/sbin/arp -d` call; we've encountered a race condition where by the time `/usr/sbin/arp -d <ip>` is called, the stale entry has already been cleaned up. This code should probably be wrapped in a `try: except:` and ignored on failure (as it's just cleanup, after all).
This bug *only affects Liberty and prior; Adam Gandelman actually fixed it as a side effect in Mitaka during his work on keepalived HA:
https:/
Fix proposed to branch: stable/kilo /review. openstack. org/317696
Review: https:/