Huge disk I/O activity due to temp files activity
Hi, For a few fays I've been experiencing a very annoying problem on one of my Exchange 2007 server (SP3 UR3, on a win2k3 SP2, system + logs on a RAID1 array, DBs on a RAID5 array, all disks are 15krpm SAS) : after everything has been running fine for a random interval of time (few hours to a few days) the store.exe process starts to create, read and write several files named ifs*.tmp in windows\temps, which quickly occupy several GB of space and completly overwhelm the disk array with I/O, wich in turns brings the entire server to a crawl. The temp files and the symptoms either disapear on their own after a while, or I have to manually dismount my mailbox information bank : after that I can see the temp files disappearing one by one, and the bank is not dismounted until the last one is gone, wich can take a while. After that it's back to square one, everything will be fine for a while and then the problem will come back. Even after upping the log levels I don't find anything usefull or any related errors in them, there hasn't been any change in the server config ou any recent update installation before the problem appeared. I found somewhere on the web someone with a similar problem, but in his case the file-level antivirus was the source of the problem but I don't have any on this server. There's no shortage or RAM or CPU time on this server, the hard drives have been tested and found without problems, so I'm a bit at a loss and I thank you all in advance for any help or suggestion with this issue.
March 21st, 2011 10:56am

Is this process happening during the store mainteance schedule?Ed Crowley MVP "There are seldom good technological solutions to behavioral problems."
Free Windows Admin Tool Kit Click here and download it now
March 21st, 2011 11:23am

No, it can happen anytime.
March 21st, 2011 11:25am

"The IFS creates these files (tmp) when a large object is streamed into the store and the .stm file is too fragmented to have the entire object written in it. For example, a large object can be a message or a file." Can you check your eventviewer and ensure that online defragmention is completing? Are you allotting enough time for Online Defrags to complete at least within 1-3 days.James Chong MCITP | EA | EMA; MCSE | M+, S+ Security+, Project+, ITIL msexchangetips.blogspot.com
Free Windows Admin Tool Kit Click here and download it now
March 21st, 2011 1:05pm

"The IFS creates these files (tmp) when a large object is streamed into the store and the .stm file is too fragmented to have the entire object written in it. For example, a large object can be a message or a file." Well I'm running Exchange 2007, there are no more stm files, I don't know it that point still applies, and what would that mean in my context. As for the defrag, it didn't complete since the problem appeared, but before that it was OK.
March 21st, 2011 1:26pm

Ahh missed that you were running 2007; the only thing for tmp files in the context of 2007 is the replay directory (pickup folder) maybe possibly from application relay servers. Can you open a copy of one of those tmp files and examine the contents and the source from where it's coming from?James Chong MCITP | EA | EMA; MCSE | M+, S+ Security+, Project+, ITIL msexchangetips.blogspot.com
Free Windows Admin Tool Kit Click here and download it now
March 21st, 2011 1:38pm

I think I found the cause, solved my problem and retroactivly identified the chain of events leading to it, and my god is that silly... A few month ago, one of our Mac user using Entourage 2008 at the time seems to have tried to send an email with an 850MB attachment and I don't know if that user did something even more misguided or if the very sucky Entourage is to blame but in the end that email was of course nether sent but 50 copies of it, 850MB attachment included, were stored localy in the drafts folder and never synced to the server. It stayed happilly that way, using a lot of space on the local Mac hard drive for nothing but not causing any trouble, until last week we upgraded that user to Outlook 2011. That new and shiny Outlook aparently didn't like seeing all those mail not properly synced, so it tried to sync them to my Exchange server : it really didn't like receiving 40GB of mail almost all at once, and could never handle all of it before the user closed his Outlook or went offline or I was forced to unmount the database, so everything restarted from scratch and the symptoms reappeared each time that user's Outlook went back online. After deleting these emails on the Mac everything seems to be back to normal.
March 22nd, 2011 7:44am

You may be running into an old issue, it's a looping bug in entourage, it was one of the common issue with rapid transaction log growth. How to allow Entourage for Mac to send large e-mail messages to an Exchange 2007 server http://support.microsoft.com/kb/935848James Chong MCITP | EA | EMA; MCSE | M+, S+ Security+, Project+, ITIL msexchangetips.blogspot.com
Free Windows Admin Tool Kit Click here and download it now
March 22nd, 2011 1:02pm

Thanks, good to know even if we're transitioning away from Entourage, it will not be missed.
March 22nd, 2011 1:11pm

This topic is archived. No further replies will be accepted.

Other recent topics Other recent topics