[dmcdermo]

I run a 12 domino mail server shop with 4 bes servers. When the bes servers are up and running, my domino servers spike up to 100 % (individual servers may vary within 10%).

As soon as I shut down the bes servers, the cpu on domino servers goes down to 10-30%.

bes version 4.1.6 mr 4

domino 8.5 or 8.02 (mixed, all domino servers regardless of domino version are affected.

Thanks

[CO_BBTechie]

Do they stay at 100% utilization? You may need to review your replication schedule. I suspect that's what's happening in the background. If this is your issue, the CPU should come back down after replication is complete. Especially if this is your initial replication.

[dmcdermo]

stays at 100% all day.

Once the mail delivery volume decreases (in the evening), the cpu util decreases, but still remains far above 'normal'. Only way to get cpu down to normal is shut down bes.

[Aroc]

1. What Antivirus are you using? Open File backup software/Agent?
2. What mail template are you using?
3. Are the BES users' mail files Full Text Indexed?
4. Any BES users have "large" mail files (>=1000 documents in the $Inbox view, >=500 Folders)?
5. Is there one specific task on the mail servers (nserver.exe? nrouter.exe?) that is consuming the most CPU time when the symptoms occur?

-Aaron

[Jadey]

How many BES users on each domino server? Are they split across the 12?

[dmcdermo]

Thanks for taking the time to read and post. I appreciate your help. Answers listed below:

1. our a/v is disabled. Tivoli open file is not running during the day.
2. we have 3 mail templates: 8.5 (rolling out new 8.5 client, not many users), 6.55 (majority of users), 6.55 Salesplace (very small # of users)
3. no
4. we have some very large mailfiles which exceed the #'s you listed.
5. nserver.exe is the task that is taking all the cpu

6. bes users are distributed across all our mail servers.

we upgraded our bes servers to 4.1.6 mr 4 about 2 months ago, unfortunately I don't have historical cpu stats that can tell me if this upgrade was the culprit or not. Originally we we had mr2, then had seperate updates to mr3 and then mr4.

[Aroc]

Hello -

I haven't tried the R8 mail templates as of yet, that's why I asked. The majority of our users are still on the 6.5.6 template. In speaking with both RIM and IBM over the past few years, both report CPU and memory utilitization on the Domino mail servers and BES servers will be greater depending on certain thresholds. Though the range I'm talking about is 10% versus 40%, not 100%. The numbers I cited in my post above we from RIM themselves. I, too, have users that exceed 1000 documents in the $Inbox view or 500 folders, and I don't see the CPU pegged to 100%.

One other perfmon metric you might watch is Avg Disk Queue Length. I've seen people suggest keeping the Avg Disk Queue Length value under the number of spindles in that array. Though even numbers as high as 2x the number of spindles is probably OK. If this number is constantly pegged high, you will probably see other perfomance problems as a result.

I have a couple tests you might try.

1. temporarily stop and disable the BlackBerry Synchronization Service windows service. This controls wireless contacts/tasks/memos/user_device_settings synchronization. The sync service adds a fair bit of load, and this might help isolate it as the culprit.

2. temporarily stop and disable the BlackBerry Alert service. Sometimes this service goes haywire when probing mail files to gather statistics.

I've also seen higher CPU load as a result of some issue with wireless Calendar reconciliation. I believe there have been various SDRs logged by RIM on these that were addressed in the various MRs. The problem with calendar issues is they are almost impossible to track down unless you luck out and get a user to come forward reporting problems with his or her calendar. I've seen this cause higher CPU consumption for extended lengths, however, I've never seen it peg the CPU to 100%.