Topic
  • 4 replies
  • Latest Post - ‏2013-01-22T20:37:17Z by Bill_Buros
Ozan_Uzun
Ozan_Uzun
5 Posts

Pinned topic Weird CPU output, ping latency on Redhat 6.3

‏2013-01-17T09:37:32Z |
 Hi,
 
On 1 socket PL server, ( firmware AL..95, installed on VIO/IVM, 3 Lpar's , Redhat 6.3 , 27G of dedicated RAM, 1/3/4 CPU, running websphere 7.x ) we are having issues.
 
First of all, the the average output of systat packages ( mpstat,iostat,vmstat)  are not as is supposed to be.
Eventhough one of the threads are busy, (11 are free) the average output is 100% usr, so the sar output is always > 90% , because the average parameter is wrong.
 
Also when the load on the server decreases, and the entitled cpu drops 0.1 and below, the ping latency of the server jumps to 300-1000 ms, when the load increases, the ping latency is ok again.
 
Is there a fix, workaround for this?
 
Updated on 2013-01-22T20:37:17Z at 2013-01-22T20:37:17Z by Bill_Buros
  • Bill_Buros
    Bill_Buros
    167 Posts

    Re: Weird CPU output, ping latency on Redhat 6.3

    ‏2013-01-17T12:56:12Z  
    For the CPU metrics, yes, that's a known problem.   Jenifer Hopper nicely documented the fix (you need the latest zstream update) and elaborated on the problem here.   The original problem was reported here.
     
    I'm not sure what you mean with the ping latency and the entitled CPU.     Is this an issue with a relatively dormant partition trying to wake up?   (my words..)
  • Ozan_Uzun
    Ozan_Uzun
    5 Posts

    Re: Weird CPU output, ping latency on Redhat 6.3

    ‏2013-01-17T13:45:34Z  
    For the CPU metrics, yes, that's a known problem.   Jenifer Hopper nicely documented the fix (you need the latest zstream update) and elaborated on the problem here.   The original problem was reported here.
     
    I'm not sure what you mean with the ping latency and the entitled CPU.     Is this an issue with a relatively dormant partition trying to wake up?   (my words..)
     Thx for the info, the  cpu usage meter is working properly now.
     
    Ping latencty issue is repeatable,
     
    when the phy. cpu usage on the server drops to  very low values and recovers after that,  in that moment ( a couple seconds actually) the response time of the server ( actually ping latencty) jumps from <1ms to
    300 ms and above. Will try with the newer kernel now.
     
  • Ozan_Uzun
    Ozan_Uzun
    5 Posts

    Re: Weird CPU output, ping latency on Redhat 6.3

    ‏2013-01-22T16:51:34Z  
    • Ozan_Uzun
    • ‏2013-01-17T13:45:34Z
     Thx for the info, the  cpu usage meter is working properly now.
     
    Ping latencty issue is repeatable,
     
    when the phy. cpu usage on the server drops to  very low values and recovers after that,  in that moment ( a couple seconds actually) the response time of the server ( actually ping latencty) jumps from <1ms to
    300 ms and above. Will try with the newer kernel now.
     
     The root cause has been found. There was an issue, virtual ethernet / LPAr in IVM.
     
    When we disabled the Virtual Ethernet from IVM, and after that provisioned the physical ethernet  card to the LPAR profile, the ping latency problem is solved.
  • Bill_Buros
    Bill_Buros
    167 Posts

    Re: Weird CPU output, ping latency on Redhat 6.3

    ‏2013-01-22T20:37:17Z  
    • Ozan_Uzun
    • ‏2013-01-22T16:51:34Z
     The root cause has been found. There was an issue, virtual ethernet / LPAr in IVM.
     
    When we disabled the Virtual Ethernet from IVM, and after that provisioned the physical ethernet  card to the LPAR profile, the ping latency problem is solved.
     Thank you for posting the answer..