[
https://issues.apache.org/jira/browse/DIRSERVER-2093?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14747086#comment-14747086
]
Emmanuel Lecharny commented on DIRSERVER-2093:
----------------------------------------------
Hi John,
there is a know issue with teh backend we currently use (JDBM) which may be
corrupted when some concurrent modifications occur. There is no way we can fix
this in JDBM, so we are implementing a new Backend (Mavibot) which will solve
this issue. Mavibot is a MVCC based backend, which means the database will be
eventually consistent.
In order to mitigate the issue, we have added a lot of locks around the updates
in the current code, but it's not that easy and it slow down the updates a lot.
Mavibot is pretty advanced, but we still have to implement the transaction
system which allows updates across many B-trees in one transaction. We just
discussed some aspects of it this week-end, and we intend to work on that in 2
weeks.
> ApacheDS M19 got stuck and no longer returns all entries
> --------------------------------------------------------
>
> Key: DIRSERVER-2093
> URL: https://issues.apache.org/jira/browse/DIRSERVER-2093
> Project: Directory ApacheDS
> Issue Type: Bug
> Affects Versions: 2.0.0-M19
> Reporter: John Peter
> Priority: Blocker
>
> We are using ApacheDS M19 on DC1 and DC2. DC1 is primary and all editing is
> done to DC1.
> We use a custom replication scheme to push updated to DC2 once an hour.
> (MultiMaster replication has caused us issues...)
> DC2 apacheDS stopped responding with the error message:
> INFO | jvm 1 | 2015/09/12 03:28:58 | Exception in thread
> "pool-2-thread-341" Exception in thread "pool-2-thread-528" Exception in
> thread "pool-2-thread-802" Exception in thread "pool-7-thread-1716"
> java.lang.OutOfMemoryError: GC overhead limit exceeded
> This happend when the custom build replication tried searching for all the
> entries with a query: (objectClass=*).
> On DC1 it returns 28469 entries. On DC2 it received a timeout.
> After restarting DC2 apacheDS it seemed to startup normally but using
> Directory studio to search for all entries it seems to return only 6665
> entries.
> Under ApacheDS\instances\default\partitions
> DC1 we can see a size of 778MB. The master.db is 329MB.
> DC2 we can see a size of 2,56GB. The master.db is 2,46G.
> Might the DC2 apacheDS database be corrupt? Any idea why it is so large?
> Any suggestions how to recover and how to avoid the issue from reoccuring?
> We had similar issues in August and it was solved by reinstalling the DC2
> apacheDS from an ldif dump.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
