New race condition exposed when cleaning up floating ips on router delete
Bug #1373100 reported by
Carl Baldwin
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
neutron |
Fix Released
|
Critical
|
Armando Migliaccio |
Bug Description
The patch that cleans up floating ips on router deletion [1] has triggered a race condition that causes spurious failures in the dvr job in the check queue. Reverting this patch [2] has shown to stabilize it.
[1] https:/
[2] https:/
Changed in neutron: | |
milestone: | none → juno-rc1 |
description: | updated |
Changed in neutron: | |
importance: | Undecided → Critical |
tags: | added: l3-dvr-backlog |
Changed in neutron: | |
assignee: | Rajeev Grover (rajeev-grover) → Carl Baldwin (carl-baldwin) |
Changed in neutron: | |
assignee: | Carl Baldwin (carl-baldwin) → Armando Migliaccio (armando-migliaccio) |
Changed in neutron: | |
status: | Fix Committed → Fix Released |
Changed in neutron: | |
milestone: | juno-rc1 → 2014.2 |
To post a comment you must log in.
Some notes that I got from Rajeev:
Going through couple of logs it appears that the patch exposes a race condition. It looks like a delete router (A) message to L-3 clears up the fip namespace and the rtr2fip links while the L-3 is in the middle of configuring a floating IP for another router (B).
When (B) first checks for the presence of agent_gateway_port all appears good but by the time (B) comes around to configure the rtr2fp or fp2rtr, (A) has cleared up the agent_gateway_port and associated links. Causing a failure.
...
Thanks,
-Rajeev.