Marvin, thanks for clarification.
It seems the problem happens when the cleaner is running.
I'm on MySQL 5 (RHES).

We need to use jpa because of rememberMe.


Robert

Am 25.05.2011 um 14:55 schrieb Marvin Addison <[email protected]>:

>> One of my CAS Servers (3.2.1) is having a problem with deadlocks.
> 
> It's a known issue and appears to be caused in part by the design of
> JpaTicketRegistry.  The root problem is lock escalation caused by
> parent-child relationships between TGT and ST tables.  An example flow
> that illustrates the problem:
> 
> 1. Obtain read lock on parent TGT row
> 2. Insert ST with FK reference to TGT row
> 3. Upgrade lock on parent TGT to update lastUsedTimestamp and other metadata
> 
> The problem is at 3.  While this flow doesn't make clear how a
> circular lock condition is obtained, I'm fairly certain the root cause
> is lock escalation.  Additionally, the locking details are entirely
> platform specific.  Oracle handles this the best, followed by MSSQL,
> with PostgreSQL coming in distant third.
> 
> I hope to address these design issues for CAS 4, but you should try to
> triage the problem for now.  I'd recommend testing additional indices
> or dropping the FK reference between TGT and ST -- that helped in the
> case of PostgreSQL.
> 
> M
> 
> -- 
> You are currently subscribed to [email protected] as: 
> [email protected]
> To unsubscribe, change settings or access archives, see 
> http://www.ja-sig.org/wiki/display/JSG/cas-user

-- 
You are currently subscribed to [email protected] as: 
[email protected]
To unsubscribe, change settings or access archives, see 
http://www.ja-sig.org/wiki/display/JSG/cas-user

Reply via email to