DX NetOps

 View Only

  • 1.  Disk Latency Explaination

    Posted Feb 23, 2023 04:44 PM

    Hello All,

    we have a mayday situation in our environment. From past 10 years my client is using ehealth sysedge to monitor their servers and one of the metrics monitored is Disk latency. This latency value as per the trends remains below 10 ms However, yesterday suddenly so many SQL servers started throwing latency alerts for all of their hard disks. As per the windows team and app owners, there is no such issue(checked at server level as well as VROps tool) but ehealth continues to show latency above 40 ms for these servers.

    Has anyone faced this issue before? Or if you can point to the calculation of this metric so that we can be sure of data reported by ehealth ?

    Any help would be very useful.

    Thanks

    AK



  • 2.  RE: Disk Latency Explaination

    Posted Feb 24, 2023 04:50 AM

    I did not see any updates on SysEDGE with version 5.9.2.

    What OS are you monitoring? You should be trying to validate the data from multiple sources. As SysEDGE most probably relies on system calls to OS for those values, these might have been changed over the years with the introductions of newer os versions. 

    If everything is running fine on the server side (at least this I understand for your statement), changes like that don't occur in data reported unless something changes in the server side, maybe OS side. 



    ------------------------------
    Cătălin Fărcășanu
    Senior Consultant
    SolvIT Networks
    ------------------------------



  • 3.  RE: Disk Latency Explaination

    Posted Feb 24, 2023 10:39 AM

    Hi Catalin,

    Thanks for your response. Problematic servers are a mix of 2008 and 2016 as well. 

    Now from server and VRops perspective, teams have checked metrics like avg read/sec, avg write/sec but nothing alarming. Now we dont know what is combination ehealth is using to finally calculate the "Disk Latency" value.

    From server side, No activity has been performed or no changes done at all.

    Thanks

    Anmol




  • 4.  RE: Disk Latency Explaination

    Posted Feb 24, 2023 11:27 AM

    I assume you know that the average reads/sec and average writes/sec don't have anything to do with latency. As you can see, the latency is measured in msec and the values you talk about don't have any relevance in delay computation. They're a measure of load, maybe volume and not  of delay. 



    ------------------------------
    Cătălin Fărcășanu
    Senior Consultant
    SolvIT Networks
    ------------------------------



  • 5.  RE: Disk Latency Explaination

    Posted Feb 27, 2023 09:36 AM

    Hi Catalin,

    how is disk latency being calculated here ? Can you point me to the right direction please.

    Thanks