Post reply

The time stamp counter of CPU on scheduler id 12 is not synchronized with other CPUs.

  • June 11, 2007 at 9:36 am

    #172664

    The time stamp counter of CPU on scheduler id 12 is not synchronized with other CPUs.

    Does anyone have any experience with this message?  This message appears in the SQL Server log every hour accross 3-4 diff scheduler ID's and it changes from 2 - 13.

    I've done some research and I'm familar with the following.

    http://blogs.msdn.com/psssql/Default.aspx?p=2

    http://support.microsoft.com/kb/931279

    This is on a 2 node Active/Passive cluster w/ Microsoft Certified Cluster Hardware.  HP ProLiant DL580 G3

    I've done everything on this KB page and its still happening.  I've done a bunch of work with the server to ensure that there are no power management features enabled that are changing the CPU freqs.

    OS: Power:Always ON

    ILO: Power Regulator: Disabled

    BIOS: HP Dynamic Power Savings Mode : Disabled

    Is this message really something to worry about?

  • July 11, 2007 at 12:28 pm

    #718747

    Hello Andrew,

    I am seeing the same message in our logs. Did you ever find a good answer?

     

    Thanks,

    Keith

  • July 11, 2007 at 12:39 pm

    #718748

    Microsoft makes it out to be a hardware problem or a software setting (example: Power Management set to always on) that affects the way the hardware functions.

    So…

    I've gone rounds with HP (the hardware vendor) and they have had me check and change a number of settings from ILO Power Management Settings, BIOS Updates etc. and nothing seems to make a difference.

    It doesn’t appear to be causing any serious problems it’s just more of a nuisance right now so this issue has received the back burner.  I’m still interested in resolving this and if you would like to compare notes let me know.

  • July 11, 2007 at 12:48 pm

    #718751

    Thanks for the update. Our server admins had asked our SQL DBA's to look into this so I have been doing some research with out much luck either. It appears it may be something with the server settings. I am going to volley this back to the server admins so they can work with Dell. I will post something back here if we find a good resolution.

    Thanks again,

    Keith

     

  • July 11, 2007 at 1:39 pm

    #718766

    I have been through this myself and the solution was to Turn HT OFF on HP servers based on Xeon Procs!

    Hope this helps.


    * Noel

  • July 11, 2007 at 1:45 pm

    #718770

    Correct me if I'm wrong but wouldn't that cut your potential processing power in half?

  • July 11, 2007 at 1:54 pm

    #718774

    Not exactly accornding to *many* people HT can potetially be even BAD for DB Servers. And for App Servers the gain could only go up to 30% more (remember these are "logical" CPUs not *physical* ones)

    If you need confirmation for my First statement, here you go:

    1. http://blogs.msdn.com/slavao/archive/2005/11/12/492119.aspx

    2. You can also search on this site for several forums about HT

    In *all* cases the real answer is TEST,TEST,TEST... ymmv


    * Noel

  • July 11, 2007 at 2:03 pm

    #718777

    Thanks noeld!

    I also just came accross this.

    http://news.com.com/Does+hyperthreading+hurt+server+performance/2100-1006_3-5965435.html

    and it links to the first item in noeld's post.

    I'm going to test one node with out HT and then failor and test on the node with HT and I'll report back.

  • July 11, 2007 at 2:28 pm

    #718785

    Great. I'll be waiting for your reply


    * Noel

  • July 12, 2007 at 6:29 am

    #718944

    Most of the websites and blogs I am finding about this are from late 2005. Have either of you seen any more up to date information? The reason I ask is we have a 64 bit box with four dual core 3.4 Ghz Xeon cpu's running SQL 2005 Enterprise. After doing some research I thought we would be better with HT off then I saw Slava made a reply at the bottom of his blog http://blogs.msdn.com/slavao/archive/2005/11/12/492119.aspx 

    Below is Slava's response to a user on 11/29/05.

    Q1: What implications does this have for dual-proc machines?

    Q2: From your explanation I take it that multiple processors or multiple cores should not show the same problem because of the lack of shared L1 and L2 cache?

    A12: That is correct. Since dual procs don't share caches we shouldn't see this behavior. The same stands for cores that share neither L1 nor L2.

    Unfortunately we only have the one server that has the dual cores and it is already in production. I need to have some good evidence to convince our Sys admins to make any changes to the server configuration. Any help would be greatly appreciated.

     

    Thanks,

    Keith

     

     

  • July 12, 2007 at 8:10 am

    #718975

    Dual core is *NOT* the problem HT *IS*


    * Noel

  • July 17, 2007 at 12:50 am

    #719890

    I'm getting exactly the same error messages from my Itanium server. No power utils are running on th box. It's single-core, none-HT.

    Any ideas as to the cause, or should I take it up with HP?

    Gail Shaw
    Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
    SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability

    We walk in the dark places no others will enter
    We stand on the bridge and no one may pass

  • July 17, 2007 at 8:51 am

    #720070

    Disabling hyper threading appears to have resolved this issue as I’m no longer receiving the warnings in the logs.  The average % processor time seems to be the same so it doesn’t appear to have changed the performance.  I'm going to spend some more time testing and such.  Whats nice about my situtation is I have another passive node with hyper threading enabled that I can use to test with =)

     

    Thanks Noeld!

  • July 17, 2007 at 11:13 am

    #720110

    Glad I could help!

    I did spent *some* time on this and at least it saved you some grief


    * Noel

  • June 17, 2008 at 9:29 am

    #829886

    Is it safe to ignore this event?

    I added the flag /rpmtimer and updated the AMD proc driver for the opteron and the event disappeared for a couple of days but it came back all of a sudden and i´m running out of ideas.

    Maybe setting the affinitty of CPU in the SQL properties page?

    Anyone found a fix or workaround?

Viewing 15 posts - 1 through 15 (of 26 total)

You must be logged in to reply to this topic. Login to reply