jakarta-jcs-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aaron Smuts <aasm...@wisc.edu>
Subject RE: Additional problems.
Date Mon, 26 Apr 2004 00:10:02 GMT
> >
> > I feel like I missed the first part of this conversation.
> >
> > So you have the updated code?
> Well I uncommented out the code that was clearly mean't to
> be in there. The part where the remote server recieves a
> "put" and then iterates over all the clients and sends them
> remove commands. I can do a diff check if there is any other
> patches since I recieved the code.

I meant, do you have the code from cvs with the patches from last week.
I'm not sure what commented section you are talking about.  

>   > Can you paste your configuration files into an email.
> Apart from uncommenting the callback ports they haven't
> changed much, on monday I will repost the remote and normal
> ccf files.

Are you running the memory shrinker?  If so turn it off and see if
things improve.  

Are the local caches and the remote cache using a lot of memory, just
the locals, or just the remote?

> >>The data we are caching comes to about 10Meg maybe 20 at most and
> >>remotecache was running with 350Meg memory!
> >
> >
> > What kind of stuff are you caching?
> Binary data byte[]. Mainly this binary data is actually HTML
> pages probably varying in size between 4k and 10k but some
> of the binary data is image data (gifs). both are put in
> simply as byte[]'s

> It is using the disk cache, however we only cache around
> 3000-4000 items. I increased the memory cache size to 4000
> as I thought the whole cache should fit easily into memory.
> Since I did this the disk cache remains empty. FYI when the
> memory cache was size 10 the size of the disk cache was
> about 10meg.

> All the machines have the same items which only come to
> around 10meg in size. 2 machines are generating updated
> content and frequently putting new data into JCS to replace
> old (with the same keys of course). So around 60% of the
> data in the cache is overwritten with new data every
> 15minutes or so. This is because we process realtime-ish
> data and need to update the html in the cache to reflect
> this data.
> Even if items were never expiring based on time then it
> shouldn't be a big problem as new items are rare.

So you are using the shrinker.  Turn it off.  I think your old ccf file
was in your message.  It had the maxlife set to 7 seconds or something
very low and the shrinker on.  If you have constant updates, just let
the updates replace the old content and anything that expires die when
it is queried for now.


Let me know if killing the shrinker helps.  Something must be keeping
references to the discarded items.


To unsubscribe, e-mail: turbine-jcs-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: turbine-jcs-user-help@jakarta.apache.org

View raw message