Last modified: 2014-07-21 10:55:27 UTC
when Gblocking: I got: A database query error has occurred. This may indicate a bug in the software. Function: GlobalBlocking::insertBlock Error: 1062 Duplicate entry '209.126.72.83-0' for key 'gb_address' (10.64.16.22) input from IRC: <legoktm> matanya: yeah race condition, file a bug for it please? :) <csteipp> It does a check, then does the insert, so definitely a race condition.
(In reply to matanya from comment #0) > <csteipp> It does a check, then does the insert, so definitely a race > condition. [19:08:34] <legoktm> well, the check we do before hand uses the slave, not master
Our check to see if there is already a block is currently using a slave. Instead it should be using master. Then, I'm thinking we should set 'IGNORE' on the insert query, and if we didn't change any rows due to a block already existing, display an error indicating so.
Change 145503 had a related patch set uploaded by Legoktm: Avoid race condition in GlobalBlocking::insertBlock https://gerrit.wikimedia.org/r/145503
Change 145503 merged by jenkins-bot: Avoid race condition in GlobalBlocking::insertBlock https://gerrit.wikimedia.org/r/145503
legoktm: Patches mentioned in this report has been merged - is there more work left to do here (if yes: please reset the bug report status to NEW or ASSIGNED), or can you close this ticket as RESOLVED FIXED?
This should be fixed now :)