Launchpad itself

many bug activities timing out due to contention on bugsummary

Bug #794802 reported by Bryce Harrington
16
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Launchpad itself
Fix Released
Critical
Robert Collins

Bug Description

OOPS-1986EA9

 - bugsummary is doing table rather than row level locks
 - bugsubscriptions are causing full locks to occur on all bugs

See original description

Related branches

lp:~lifeless/launchpad/bug-794802
Stuart Bishop (community): Approve
Diff: 875 lines (+744/-8)
10 files modified
database/schema/patch-2208-63-1.sql (+460/-0)
database/schema/patch-2208-63-2.sql (+13/-0)
database/schema/patch-2208-63-3.sql (+55/-0)
database/schema/patch-2208-63-4.sql (+171/-0)
database/schema/security.cfg (+15/-5)
lib/lp/bugs/doc/bugsummary.txt (+5/-0)
lib/lp/bugs/model/bug.py (+3/-2)
lib/lp/bugs/model/bugsummary.py (+1/-1)
lib/lp/bugs/model/tests/test_bugsummary.py (+18/-0)
lib/lp/registry/model/person.py (+3/-0)
Revision history for this message
Bryce Harrington (bryce) wrote :

Also tried doing this (adding 'linux') from a different bug:

https://bugs.launchpad.net/ubuntu/+source/fglrx-installer/+bug/776895/+distrotask
OOPS-1986EB9

However, it worked when I did it from the following bug but added 'jockey' rather than 'linux':
https://bugs.launchpad.net/ubuntu/+source/jockey/+bug/792240

Revision history for this message
Bryce Harrington (bryce) wrote :

Retargeting an existing bug task to 'linux' also seems to not work: OOPS-1986AX12

William Grant (wgrant)
Changed in launchpad:
importance: Undecided → Critical
status: New → Triaged
tags: added: regression timeout
Robert Collins (lifeless)
summary: - OOPS-1986EA9 trying to add 'linux' task to a bug
+ many bug activities timing out due to contention on bugsummary
description: updated
Revision history for this message
Robert Collins (lifeless) wrote :

We're basically doing - way- to many updates to the bugsubscription table, I've a fix for that. We may need to change the locking model to handle more concurrency though.

Revision history for this message
Robert Collins (lifeless) wrote :

My attached branch makes this use row level locks rather than table-wide ones, and only modified rows which really would be modified. Its probably not perfect at this - we may still have two disparate sql statements that would null each other out, but at least on a per-statement basis it will avoid pointless row locks being taken out.

Tom Haddon (mthaddon)
tags: added: canonical-losa-lp
Revision history for this message
Launchpad QA Bot (lpqabot) wrote :

Fixed in stable r13211 <http://bazaar.launchpad.net/~launchpad-pqm/launchpad/stable/revision/13211>.

Changed in launchpad:
assignee: nobody → Robert Collins (lifeless)
tags: added: qa-needstesting
Changed in launchpad:
status: Triaged → Fix Committed
Robert Collins (lifeless)
Changed in launchpad:
status: Fix Committed → Fix Released
tags: added: qa-ok
removed: qa-needstesting
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.
Subscribing...

Other bug subscribers

Related questions

Remote bug watches

Bug watches keep track of this bug in other bug trackers.