jakarta-jcs-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Niall Gallagher <ni...@switchfire.com>
Subject JCSAdmin.jsp doesn't broadcast removes?
Date Wed, 26 Sep 2007 18:39:00 GMT
Hi,

Not sure if this has been asked before..

We have found that the JCSAdmin.jsp running on our central cache server,
does not seem to broadcast removes to client machines when we clear
regions via this JSP on the central server. Is this expected behaviour
or a bug?

We have gotten around the problem by writing our own custom JSPs based
on the JCSAdmin.jsp and JCSAdmin bean, and deploying them in the same
webapp as the JCSAdmin.jsp and jcs jar file (let's call this the "jcs
webapp"). Our custom JSPs don't actually clear regions themselves, they
POST the requests to another webapp we have written.

The other webapp we have written (let's call it "jcs assistant")
contains just a simple servlet which listens for HTTP POSTs asking it to
clear certain regions or elements. This servlet clears the requested
region using the normal JCS API as if it was a normal client machine
(note that we only use string keys, so we can send them in the POST).

Both webapps are deployed to JBoss 4.2.1GA. We found that we had to
configure Isolated -> true in the ear-deployer.xml file, to force JBoss
to load a separate copy of JCS for both webapps.

Basically- we've found that when a client puts an object in the cache it
gets sent to the central server and shows up on both our JSPs and the
JCSAdmin.jsp.
If we remove the object from the cache via JCSAdmin.jsp, it disappears
from both sets of JSPs, but if the client then tries to retrieve the
same object from the cache again it unexpectedly *succeeds* (this is
without the client shutting down between runs).
If we remove the object from the cache via our "JCS assistant" webapp,
it disappears from both sets of JSPs, and if the client then tries to
retrieve the same object from the cache again it gets null as expected.

Has anyone seen this behaviour before? Is there anything we can do to
fix it? Obviously having to hack JSPs and deploy two webapps isn't a
very good solution to this problem.

Thanks in advance,

Niall


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message