WSM on Server 2K8 R2 anyone?
We've deployed WSM on Server 2008 R2. In the central site I have two servers (core + edge) for load balancing. Lately we've boosted the number of streaming clients from 20 to 80, but this should work fine even on one server.
I have a problem with what I suspect is a memory leak somewhere. After the servers have been turned on for about a week all memory is used and they start swapping. This again leads to ENORMOUS performance problems on the clients, with boot times up to 30 minutes (worst case).
Both servers runs on vmWare ESX4 with 2 vCPU's and 4 GB RAM.
PS! The Java version that comes with WSM is v6u7. This is more than 2 years old, and I've seen memory leaks from Java before. Is there any harm in upgrading the Java version to the newest one?
Any clues will be greatly apreciated!
I have not deployed WSM on 2K8, but here is an old thread which might be of help to you:
I'm considering re-installing on Server 2003, just to check if that helps. I'm still seeing a lot of random performance issues, and I'm not sure if it's server or network related (it's got to be one of them). Boot time varies from 30 seconds to 30 minutes :S
I'm also seeing that the writecache for each client grows quite large. To me it would seem logic if the writecache is cleared when a client shuts down, since the image is streamed in volatile mode. On some occations I've deleted the writecache of a client with poor performance, and it has helped.
I'm talking about the writecache directory on the server.
I've not yet set up a scheduled task, but I've deleted the files manually for testing.
Re-installing server AND creating a new image as we speak. This way I'll find out if it helps to run it on Server 2k3. I'm also suspecting our routing setup to be a bit of a bottleneck, so if this doesn't improve performance I'll try putting a server in the same subnet as some of the clients.
I'll report my results back here.
The writecache shouldn't keep growing, it is deleted when a client in volatile mode boots on a server that already has the writecache file on disk.
However, if you patch the OS, a new writecache file is created.
A client in persistent mode is another matter.
Of course, you need enough free disk space for the OS.
You wrote that your servers are virtual. My experience is that these are generally much slower than physical servers. We usually let core and DB be virtual, and the rest physical, as this gives much better performance.
I've re-installed on 2003, and allthough the end-user performance issues are still there, the server itself behaves much better, and it seems like the writecache files aren't growing that big either. We were also seeing other issues on 2k8; services stopping unexpectedly, web interface painstakingly slow etc. This has not been the case on 2k3.
Next step is to go physical.
Have you been able to get better performance on physical platform?
We also have VMware ESX 4.1 and WS2008 R2 and are experiencing similar performance problems with ~60 R00L clients.
Monitoring storage, network and resources utilization doesn't seem to reveal any issues, but boot times are often dog slow.
Thanks in advance,
Very interesting - I think you're the first guy with this problem I've run into besides myself.
I have not yet tried it on a physical box. As a temporary measure, I've just created more virtual servers and distributed the client load across them. Performance is decent as long as the client count doesn't exceed ~40.
I've also confirmed that WSM (3.5) runs better on Server 2K3 then on 2K8R2.
I will do more testing, including using a physical server, but I haven't had the time to do it yet. When I do, I will post my findings in this thread, and it would be great if you did the same.