> Why would you want to introduce synchronization into this?

Because it will likely solve the interleaving problem Pat reported.
I'm reasonably certain my proposed explanation of the behavior is
correct:  multiple http request threads competing for IO.  This
assumes inspektr is using the request thread for writes instead of a
single worker thread with an unbounded queue.  You could easily verify
this by adding the thread ID to the audit trail file appender layout:

log4j.appender.logfile.layout.ConversionPattern=%d (%t) %-5p [%c] - %m%n

where %t will produce the thread ID in the log file.

>Clearly
> serializing didn't help (since these things are put in one queue and popped
> off one at a time <-- at least it used to be).

As far as I can tell, only JdbcAuditTrailManager works that way.

> Synchronization isn't going to do much better probably.

I'll put a tasty beverage on the line that synchronizing the record()
method will fix it.

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-dev

Reply via email to