ESXi

 View Only
  • 1.  Identify storage read latency peaks reason

    Posted Oct 05, 2016 12:02 PM

    Hi ,

    Here is a brief summary of my VMWare arch.

        

    HV :

         Dell 730

    Storage :

         QNAP TVS-EC2480-U-SAS-R2 (10 SAS DISK 10K HGST in raid 10 and 8 SSD Intel SC3600 in raid10)

         2 LUNS on SAS Raid Group and 2 LUNS on SSD Raid Group

    HV is connected to Storage by 2 10Gbps SFP+ links in MPIO. Luns are mounted in ISCSI and Jumbo frames are OK everywhere.

    I got matters with a Lun on my SAS Group. It hosts 3 VM and i often got latency read peak on the datastore.

    When i check VMS graphs (IOPS / CPU / MEM) at the peak moment i can't find any reason that could provocate it.(1200 ms Max about 10-15ms AVG).

    No problems with the Write latency and no matters with my other Luns (Read / Write)

    What could i check ? What should i provide to help this ticket ?

    Sorry for my english , i'm French

    Best regards

    Pacopad



  • 2.  RE: Identify storage read latency peaks reason

    Posted Oct 07, 2016 08:55 PM

    >i often got latency read peak on the datastore.

    Could you check on /var/log/vmkernel.log file when you try to read a file from the datastore. It need not be a VM, you can write a file to it and read it.

    If you get lots of messages related to I/O error, then please do check on the VMware kb about SCSI errors to find out if there are any locks or driver related issues. https://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1002293

    Also, please do provide the SAS driver under topic



  • 3.  RE: Identify storage read latency peaks reason

    Posted May 17, 2017 12:49 PM

    Hi Eveybody,

    I found a solution sometimes ago. It was caused my the SSD cache on my qnap TS2480.

    I disabled it and now it rocks

    Best regards

    Pacopad