Hi, I currently have a clustered RMS running in my production SCOM R2 environment. Since the environment has grown quite a bit, and we're currently also running the Savision Livemap solution on the RMS I've started seeing performance issues with this machine. The CPU regularly flatlines at 100%, with a Dual 3Ghz AMD Opteron 2222 machine (I've noticed this with long wait on Livemaps, Changes in DA's and such..) . I have the option to upgrade to a Quad CPU Quad Core Xeon X5570. I'm expecting a lot of performance improvement from this. (or am I wrong here?) However, I'm looking for feedback on the way to do the HW upgrade. I'm thinking about a 'dirty' upgrade where I remove a node off the cluster, install the new HW under the same name as the old node and rejoin the cluster. Then do the same to the other one. This would mean I don't need any trickery with SPN's and such.. But I'll need to update the machine AD-record since I will have a new machine (SID) with an old name.. I believe all the config is in the OpsDB, so as long as I bring SCOM down for the migration it will not notice the change of HW... or is there any information that SCOM has stored locally (registry?) which also determines it is (part of a clustered) RMS? I know the safe way is a demote/move the RMS to a single machine and then install a new cluster and promote it to RMS... but I'm wondering if the way I described above (or anyone have any other idea?) is a quicker / easier way to do this...? Thanks for any ideas / feedback on this! regards, Pieter

My experience of clustered RMS is you have to play it by the book So for me by the book here is, get your new server ready to go demote on of the nodes perform the clustered install register the SPN life is good On the RMS you have registery settings that tell it how to find data bases etc that is best worked from the rms demote commandPaul Keely

Why don't you just add the new nodes to the cluster and remove the old nodes? We support more then a two node cluster. Also, how many agents in the MG. how many consoles? robRob Kuehfus | System Center Operations Manager | Setup and Deployment Program Manager

Ok, this IS the answer I was looking for. As long as adding a node afterwards to the RMS cluster and running a 3 node cluster (even if for a while) is supported by MS I'm happy with it! Size wise the MG has 500+ MS agents, 1000+ Xplat agents, 40+ Network devices and a whole load of other snmp / oledb / synth transaction devices including SCVMM for connecting to multiple ESX 3.5 / Vshpere environments. Running on 6 MS's , 2 SQL RS's , a dedicated SQL cluster for the DB's. Currently 10- concurrent consoles, but also about 5 - 10 webconsoles (running from one of the MS's). Adding to that are the MOSS2007 SLD dashboards with Savision Livemaps installed on the RMS cluster to provide single-sign on functionality. This has increased the load on the RMS quite a bit according to the performance reports and we're ready to move it to better HW. I'll add the node and let you know my experiences.. thanks! Pieter

Can you tell which process is flatlining you at 100%? SDK? Rob Kuehfus | System Center Operations Manager | Setup and Deployment Program Manager

Rob, top 3 processes are HealthService, w3wp, SDK. It is often the combination of the 3, where when the CPU is at 100% it is especially the loading of the Livemaps that shows the lag then, where the load takes seconds to complete. What I read on the subject is that you need to keep the RMS's hands as free as possible, which is why we moved the Webconsole to a different machine. However, with the Savision Livemaps tooling we were not able to do this without forcing a user to login twice for a combined Sharepoint dashboard containing both SLD and Livemap info.. Basically the computing power of the SCOM RMS should at least grow 4x according to the HW guys.. also we'll upgrade to the new corporate 'default' machine / manufacturer so there are multiple reasons for the HW swap. Still, since you ask: would you expect this kind of hardware (the old one) to easily handle the load of the described environment? Should I worry that I have other issues causing the performance problem? (or to put it differently, that the HW swap will not fix the issue?) We have had issues with performance on the SCOM infrastructure before, but after a MS PS case we've upgraded the number of MS's, and separated the DB's over 2 servers. This solved the issues then. Thankx for any feedback on this!

I am not 100% convinced that buying new HW is going to solve your problem. Do you have any agents reporting directly to the RMS? Is the current machine 64-bit? What does your disk usage look like? Rob Kuehfus | System Center Operations Manager | Setup and Deployment Program Manager

Nope, no agents talking to RMS directly, W2K8 64 bit RMS cluster with 64 bit SQL on the SQL cluster. Disks are running OK. both the local disk queue / I/O and the LUN's that are there for the clustered disks are far from being stretched to their limits. I've been keeping an eye on the disk / memory / network / cpu load and it's only the CPU load that spikes from what I can see. No huge Disk I/O or Queue's at that point.. Which is why the increase in raw computing power looks that appealing to me.. :) We already have this HW inhouse, and it's considered a 'spare' so there are not additional costs. Also the other reason mentioned (standardization and such) are reason enough. I'd be a bit disappointed if there isn't any improvement in the performance of the dashboards though. But then I'll go to Savision and complain to them that their product has to run on the RMS for integrated security to work.. :) For the SCOM webconsole there is a workaround, maybe they will provide something similar :)