1

We have an increasing number of ActiveSync devices connecting to our Exchange 2010 infrastructure (one primary mailbox server with about 920 accounts and counting; plus a cas/hub in the primary site). Solarwinds has been frequently throwing "Component: Information Store: I/O Database Writes Average Latency" alerts. I ran the troubleshooting scripts (Troubleshoot-CI.ps1; Troubleshoot-DatabaseLatency.ps1; Troubleshoot-DatabaseSpace.ps1) and all came back as a pass/good.

Has anyone seen a correlation between increasing ActiveSync devices and database write latency? If so, how did you resolve it?

marcwenger
  • 235
  • 1
  • 6
  • 21
  • What is the average latency? – Greg Askew Jan 31 '13 at 18:59
  • On the critical alerts from solarwinds, that means average latency of over 100ms during the polling cycle of 300 seconds. We're gathering more info. Everytime we get such an alert we'll check the solarwinds console to see what the latency was recorded as – marcwenger Jan 31 '13 at 20:27
  • I don't see a number of activesync devices attributing to capacity more than an non-activesync device. Some activesync devices/versions are known to be not good network citizens, but I would not assume that to contribute significantly to database write performance. – Greg Askew Jan 31 '13 at 20:35
  • Thanks, Greg. Today the server seemed to have latencies usually below 5 ms. I guess I'll search what else could be causing latency alarms. – marcwenger Jan 31 '13 at 23:42

1 Answers1

0

I worked on an issue today where transaction logs completely filled a drive in a short amount of time due to misbehaving ActiveSync clients (ended up being iPhone 4). The IIS logs for the day were about 10x the usual size.

The script available here in conjunction with LogParser directly identified the users that were hitting ActiveSync tens or hundreds of thousands of times per day, vs. hundreds of times for everyone else.

http://blogs.technet.com/b/exchange/archive/2012/01/31/a-script-to-troubleshoot-issues-with-exchange-activesync.aspx

If you have unruly devices, that could be slamming your I/O.

Hope this helps.

Jeremy Lyons
  • 1,088
  • 6
  • 9
  • Looks like we do have some high request devices, and users with multiple devices. We'll attempt a crackdown on multiple devices – marcwenger Feb 07 '13 at 18:57
  • I just saw this over the weekend which explains the unruly IOS devices we've been seeing: http://www.zdnet.com/ios-6-1-banned-from-corporate-servers-due-to-exchange-snafu-7000011064/ – Jeremy Lyons Feb 11 '13 at 15:21
  • It is very possible. I ran a script and counted 82 iOS 6.1 devices. – marcwenger Feb 11 '13 at 16:55
  • Did any of them have abnormally high hit counts? I was surfing this weekend and read an admin stating that only upgraded profiles on IOS 6.1 seemed to have the issue, but any newly created profiles on already upgraded devices were okay. I think you might be able to just remove and reconfigure the profile on problematic devices. – Jeremy Lyons Feb 11 '13 at 17:23