We are running Crowd 2.5.1 with a local MYSQL database on our Red Hat server.
Crowd's sync to our AD has been failing with the following error:
=======
2014-11-19 21:31:54,537 scheduler_Worker-6 INFO [atlassian.crowd.directory.DbCachingRemoteDirectory] failed synchronisation complete in [ 9334ms ]
2014-11-19 21:31:54,541 scheduler_Worker-6 ERROR [atlassian.crowd.directory.DbCachingDirectoryPoller] Error occurred while refreshing the cache for directory [ 393217 ].
org.springframework.dao.IncorrectResultSizeDataAccessException: query did not return a unique result: 2; nested exception is org.hibernate.NonUniqueResultException: query did not return a unique result: 2
=======
In practice, what this means is that JIRA and Confluence get incomplete group info, and often what is missing will be certain users will not be in the jira-users group or confluence-users group, which are the groups we use for login authentication for those apps. We have run database queries such as:
SELECT id,directory_id, lower_parent_name, lower_child_name, membership_type, COUNT(*) as cnt FROM cwd_membership GROUP BY directory_id, lower_parent_name, lower_child_name, membership_type HAVING COUNT(*) > 1;
... and have found some users with duplicate entries. We removed those users from AD and deleted the dupes in the Crowd MYSQL, and that allowed the sync to complete successfully. But once we re-added those users to AD, the problem came right back.
Some of this looks like it may be related to case sensitivity:
[root@atlassian01 logs]# grep -i mahler catalina.out.2014-11-20 | tail -1
2014-11-19 21:04:09,209 QuartzWorker-1 WARN ServiceRunner [atlassian.crowd.directory.DbCachingRemoteChangeOperations] remote username [ Mahler.Brian ] casing differs from local username [ mahler.brian ]. User details will be kept updated, but the username cannot be updated
From https://confluence.atlassian.com/pages/viewpage.action?pageId=254739329 I see that:
Cause
MySQL collation for the Crowd database is set to be case insensitive.
Resolution
Do not use COLLATE utf8_general_ci (e.g. case insensitive). Use COLLATE utf8_bin (e.g. case sensitive). The Crowd schema has keys on VARCHARs which will cause you issues if you use collation ending in CI.
However, it looks like we use utf8_bin already:
[root@atlassian01 crowd]# pwd
/var/lib/mysql/crowd
[root@atlassian01 crowd]# cat db.opt
default-character-set=utf8
default-collation=utf8_bin
I note that we have a TEST instance of Crowd 2.5.1 on an Ubuntu box, and the same corruption does NOT appear in the logs (and the sync works successfully there).
My question is, can I shut down Crowd and MYSQL, and blow away (rename) the corrupt database and then recreate it, hopefully making a new, uncorrupted MYSQL database that will sync properly? Is this reasonable? Are there risks to doing this? And how exactly would I do this (what files/directories would I delete/rename)?
