I'm using WAS 4.0.1 and seeing major CPU utilization hits on the WebSphere server. The CPU utilization is gradually increasing, then after a certain amount of time it booms up to 95%. Stays there until I re-boot the server. I think its due to SimpleFileServlet. Have you guys seen anything like this? Any patches to this version of WAS. Any ideas? :roll:
I, and many others, had this problem with WAS 3.5. It was due to WAS using InstantDB for its admin database. InstantDB is a piece of crap under Windows, it appears okay for me under Linux. Anyway, I setup WAS to use MS-SQL for its admin DB and the problem never happened again. And by the way the problem would occur for us several times a day.
Hi, I've also seen a similar problem - running WAS4.0.3 under Win2K with a remote Oracle 8.1.6 as the Admin repository. We have a Server Group of three clones on identical spec machines - 2GB ram and the same CPU (I can't rememeber what it is). We install the application with the max heap size as 256MB (the recommended max for the size of ram) and when we start the server group all the machines consume minimal RAM and CPU usage. However after moderate usage after a couple of days the JVM for Websphere consumes about 1.2GB of RAM and the CPU is maxed out - basically the server grinds to a halt and needs to be rebooted. In the Adminclient messsages window we get 'Out of Memory' errors. I've been looking at the Performance Tuning guide Redbook and the 'Out of Memory' error is a symptom of a possible memory leak within the application but the same application ran successfully without the need for reboots on Weblogic6.1 on the same physical boxes. One of my main concerns is that even though we set a max for the heap size, Websphere seems to ignore it and consumes as much of the system memory as it can before it dies!!! I've looked at the garbage collection within the JVM with the -XrunpmiJvmpiProfiler but the figures seem in line with the recommendations I'm going to run a profiler over the application but otherwise I'm in the dark! Any ideas? Cheers, Steve
Just as a sugguestion, try to set the reload parameter to be false, or set the reload interval to be longer(eg, 600). I have a similar but separate question. I found that the admin console sometimes consumes more and more memory with time, then i have to close it and restart it. when i initially start the admin console, it only needs about 100 mb memory, but after a couple days, it consumes 400-500 MB. I think that the admin console jvm does not do a good job on garbage collection. how to solve this problem? I work on WAS 3.5.x with db2 as backend and apache as web server.. thanks/
Originally posted by Dallas Cowboy: I'm using WAS 4.0.1 and seeing major CPU utilization hits on the WebSphere server. The CPU utilization is gradually increasing, then after a certain amount of time it booms up to 95%. Stays there until I re-boot the server. I think its due to SimpleFileServlet. Have you guys seen anything like this? Any patches to this version of WAS. Any ideas? :roll:
Hi Bill, I do have the reload interval set to false as its a production system and I won't be dropping files in without starting the server. I think that the default for this setting is far too low. As for your Adminclient - surely you don't sit watching it for 24 hours? So why do you need it open when you are not using it? Just close it and let WAS the adminserver run behind the scenes. Also, you can avoid using the adminclient for most tasks using WSCP scripts. I don't use them (yet!) but I've read here that a lot of people use them because the adminclient (especially on 3.5) so slow. I hopw this helps, Cheers, Steve