> With RememberMe enabled with a TTL of weeks, you might get a lot of tickets 
> in the registry.

Indeed this is true.

> This could lead to a problem because the cleaner currently locks your ticket 
> registry for a long time as all tickets are deserialized.

There is no such locking explicitly in the code, and it's hard to
imagine how _any_ database platform would lock entire tables to
perform row-level deletions in a way that would appear to lock the
entire registry.

> You can get java.sql.BatchUpdateException: Lock wait timeout exceeded; try 
> restarting transaction exceptions due to this.

I believe this is simply one manifestation of a deadlock, which is a
known issue with the current JPA implementation.  Some platforms
handle deadlocks better than others; based on my experience Oracle is
among the best and PostgreSQL among the worst.

> I'm not sure if a Jira Ticket for this problem exists.

Search the CAS Jira for "deadlock" and I believe you'll find several
issues.  We've done as best we can to reduce the likelihood of
deadlocks within the limitations of the current TicketRegistry design.
 It's doubtful it will be better until there are API changes, which
are due in the CAS 4.x timeframe.

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

Reply via email to