locked
QSM.exe 100% CPU, DE migrator RRS feed

  • Question

  • Hi, I have a HP EX490 running WHS v1 with latest updates from win update+hp.

     

    I'm seeing QSM.exe take 100% cpu for a significant amount of time - it seems as if it's about 45s out of every minute. Whilst the system appears to be functioning ok, it is a tad slow & is getting warmer/consuming more power than required., This has been going on for a long time (months?)

     

    I checked the WHS logs and the only interesting stuff I could find was in the DEmigrator log where I found errors similar to

    [12/26/2010 9:32:08 AM 2208] ***ERROR: DEfileFromShadow(c:\fs\1ID\DE\shares\Photos\P100\Germany July 2008 TJ school trip\DSC07533.JPG,9b10568a-4ddf-42ac-8920-a29e17416387) took too long. 3130.1453ms


    I went through each of these files, moved them & restarted the service

    The end result was I just get errors on different files.

    All regular access is via remote access/local UNC paths. It's possible the odd file could have been touched loclly under d:\shares -- for example using process manager I can see twonky, and the HP media collector have these paths open.

    I have tried closing off all unusual processes/services including all media streaming & this problem continues.

    Any idea how to debug/resolve?
    I'm also concerned at whether I do have files duplicated or not (though all are backed up on local uSB drive *and* remotely). I've not noticed any data access problems and no disk errors are reported.

    Is there a DE "sanity check" tool that will report on whether data is correct & check sanity - like chkdsk for normal drives (chkdsk runs clean BTW)

    Is there a way of enabling a higher level of debug/trace for DE or QSM - I'll happily take a perf/space hit if it helps identify the problem?

    Possible worst case is re-installing. Whilst I could do this I'm also concerned about the data intactness....

    Sunday, December 26, 2010 10:40 AM

All replies

  • This is driving me nutty, and worryingly there seems to be many reports of people getting it, and no resolution other than disabling drive extender (on a schedule)

     

    The behaviour I'm notice is that QSM seems to wake up from waiting on a notify very regularly -- every minute or so, and then consumes 100% cpu for 30 seconds or so as it searches through every single file in any share on my disk.

     

    So it looks as if it's hooked on the notify waiting for a "change" - perhaps a new file that might require drive extender to be involved. If so why does it then go and query every single file? And how do I find out which file caused the wakeup.

     

    I'm currently running chkde from the old server toolkit -- this is reporting shadow errors on every file which I assume is a 1.0->1.1 different, so will ignore but check for other interesting errors after completion.

    Am I expecting too much from WHS? My first thought was to either a) reinstall the server software b) forwmat & restore all data, but I see a number of people saying they've done this and it makes no difference!

    Are there any configuration options? 

    I see entries like "DirectoryWatch" in Drive Extender registry settings (I've not changed anything here). would this change behaviour?

    What is meant to happen with WHS? I have 100s 1000s files, but wouldn't regard that as a lot at all for a server. I just have two disks 1Tb+1.5Tb. Some volumes have redundancy, some don't.

    Monday, December 27, 2010 2:00 PM
  • A look through the qsm.exe log shows that I get this almost consistently every 80 seconds

     

    [12/27/2010 12:50:38 AM  a18] Fix volume snapshots

    [12/27/2010 12:50:38 AM  a18] Monitoring thread Calculating sizes

    And I think a full scan occurs following this, taking perhaps 40s leading to a "sawtooth" CPU graph.

    Monday, December 27, 2010 2:21 PM
  • I took a "process monitor" trace to see what files were being touched.

     

    There is a regular low volume of activity from qsm.exe doing a few registry queries and logging, but the huge workload comes from a scan of all folders.

     

    Just prior to this I can see

     - a flurry of activity from lsass.exe (registry stuff)

     - VSS engaged in "doing something"

     - some rpcss activity & mqsvc.exe

     

    I then checked which processes had files open on d:\shares and it was really - Twonky Media, HP Mediasmart, Windows media connect -- all apps preloaded on the EX490. I've stopped all for now to see effect

    Monday, December 27, 2010 3:17 PM
  • Twonky is probably the culprit. It does huge amounts of I/O indexing, and that can result in performance issues.
    I'm not on the WHS team, I just post a lot. :)
    Monday, December 27, 2010 4:10 PM
    Moderator
  • Thanks, though further investigation has shown IIS to be the cause. I only have the basic two sites (regular web ui & port 9000), but with EITHER running I get the high CPU prolonged spikes every 80s.

     

    Can anyone think whast in IIS repeats every 80s by default?

     

    Disabled service and issue has gone away.

    QSM will still peak occasionally as will demigrator leaving me a little concerned at the WHS design and how scalable it is especially with significant write workload. Comments?

    However the immediate issue is mitigated... still need to figure out how to have IIS working again, even though I rarely use it. Checking everything is stable for now.

     

    Note twonky does take cpu when started (updating index) but then settles - no ongoing problem with it, and also one of the few capabilities I use extensively for DLNA streaming

    Monday, December 27, 2010 8:28 PM
  • Are you sure it isn’t Twonky causing the problem? Twonky consistently uses 100% CPUI on my EX485 even though I didn’t have any media anywhere on the server. The last time I tried to get it to work, I let it run for two weeks and it was still using 100% CPU in spikes like that, so I finally permanently disabled it.
     
    I also can’t get Media Collector to work right on the EX485, so I killed it as well. The first time it collected about half my music, reported a missing file in the FTP server, and wouldn’t collect anything after that. I’ve since reset the Media Collector and it just refuses to pick up anything at all. I don’t consider it much of a loss, though. That stuff just never worked right out of the box, so I disabled all the HP junk and replaced it with a robocopy job.
     
     
    "planetf1" wrote in message news:9c27457a-f7ad-4a4c-abc0-42ebf2266620...

    Thanks, though further investigation has shown IIS to be the cause. I only have the basic two sites (regular web ui & port 9000), but with EITHER running I get the high CPU prolonged spikes every 80s.

     

    Can anyone think whast in IIS repeats every 80s by default?

     

    Disabled service and issue has gone away.

    QSM will still peak occasionally as will demigrator leaving me a little concerned at the WHS design and how scalable it is especially with significant write workload. Comments?

    However the immediate issue is mitigated... still need to figure out how to have IIS working again, even though I rarely use it. Checking everything is stable for now.

     

    Note twonky does take cpu when started (updating index) but then settles - no ongoing problem with it, and also one of the few capabilities I use extensively for DLNA streaming

    Saturday, January 1, 2011 6:52 AM