Deadlock found when trying to get lock; try restarting transaction

[Davanum Srinivas (DIMS)]

Example URL:
http://logs.openstack.org/31/107131/1/gate/gate-grenade-dsvm/d019d8e/logs/old/screen-n-api.txt.gz?level=ERROR#_2014-07-17_20_59_37_031

Logstash query(?):
message:"Deadlock found when trying to get lock; try restarting transaction" AND loglevel:"ERROR" AND build_status:"FAILURE"

32 hits in 48 hours.

[Jay Pipes (jaypipes)]

Pavel has a partial fix for the underlying problem here: https://review.openstack.org/#/c/106367/

Basically, the use of SELECT FOR UPDATE can be combined into a single UPDATE statement and reduce (most likely eliminate) the chance of deadlock exposure.

It's a partial fix because the allocate also would need to be fixed.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/106367
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=b8fe7b13cff3170cb1b87c5a26fbc8a8ec9422cb
Submitter: Jenkins
Branch: master

commit b8fe7b13cff3170cb1b87c5a26fbc8a8ec9422cb
Author: pkholkin <email address hidden>
Date: Fri Jul 11 16:54:38 2014 +0400

    Optimize db.floating_ip_deallocate

    'select for update' and 'update' operations
    were combined into a single 'update' operation

    problems with 'select for update' relate to
    spec https://review.openstack.org/#/c/97310/1

    now 'floating_ip_deallocate' function returns
    number of updated rows in db

    Partial-Bug: #1343613

    Change-Id: I60f5a8f3e1541983dea1589783927107c00c5fa4

[Matthew Treinish (treinish)]

I'm not seeing how this is a bug in tempest, besides the fact that a tempest run triggered it. Can you elaborate this is a bug in tempest?

[Joe Gordon (jogo)]

not seeing any hits for this in nova logs

[Yaroslav Lobankov (ylobankov)]

It looks like this issue is not related to Tempest. As Matthew said, tempest runs just triggered it. So moving the bug status to "Invalid".