1200-1400. We typically use a 512MB heap and our session timeout is 30 mins.
(Would be way longer if the customer had their way).
Yes, we monitor the metrics. We are not seeing any unusual spikes in the
session count before GC starts failing. On one of our lower-traffic sites,
lowering the session timeout has all but cured the problem.
We are going to look at how we can off-load some of the baggage in our
sessions, but I was wondering if those session-swapping elements might help
act as a pressure-relief valve and prevent heap memory overload.
thanks
-Chris
On Wednesday 22 September 2004 10:19 am, Michael Collins wrote:
> Chris,
>
> Everything like this needs to be tested but I don't see it causing a
> negative impact if your servers are already getting hammered.
>
> How many sessions are you talking about, are you monitoring these with
> JRun metrics?
>
> I have seen session run-ups from HTTP probes, since they do not send in
> cookies, and they run every x seconds. You might be seeing something
> similar.
>
> Mike
>
> http://livedocs.macromedia.com/jrun/4/descriptordocs/jrun-web/jrun-web.htm
> <http://livedocs.macromedia.com/jrun/4/descriptordocs/jrun-web/jrun-web.htm
>>
>
> -----Original Message-----
> From: Chris Boyce [mailto:[EMAIL PROTECTED]
> Sent: Wednesday, September 22, 2004 10:37 AM
> To: JRun-Talk
> Subject: Sessions filling up memory
>
>
> During peak hours, we have a lot of sessions resident in memory, and
> occasionally JRun grinds to a halt with failed garbage collection.
>
> Is it possible to use session elements like session-swapping,
> session-swap-interval, and session-max-resident without having to use
> persistence?
>
> If so, is there a big performance hit in JRun when swapping excess
> sessions to disk?
>
> thanks
--
Chris Boyce
[Todays Threads] [This Message] [Subscription] [Fast Unsubscribe] [User Settings] [Donations and Support]
