locked
Why will Mutlipath I-O test (LOGO) of Windows2016TP5+HLK fail? RRS feed

  • Question

  • I am executing the HLK test of the 3rd party DSM driver now on Windows2016TP5.

    Multipath I-O test(LOGO) outputs the following log, and failure continues.

    ==================================================
    (i) Begin Link Bouncing Test
    (i) Do failover on chosen path...
    (i) cmdLine is: devcon disable "@PCI\VEN_10DF&DEV_F100&SUBSYS_F10010DF&REV_03\4&2b4787d&0&0020" .
    Error: 0x37, The specified network resource or device is no longer available.
        (x) Failed to send I/O [mode = 0x00], GLE: 0x00000037.
        File=drivers\storage\tests\mpiotest\logotest\test\ioapplication.cpp Line=431
    Error: 0x0, Error 0x00000000
        (x) Normal I/O did not complete successfully.
        File=drivers\storage\tests\mpiotest\logotest\test\ioapplication.cpp Line=493
    (i) Number of I/O completed: 15543
    (i) Do failback on chosen path...
    (i) cmdLine is: devcon enable "@PCI\VEN_10DF&DEV_F100&SUBSYS_F10010DF&REV_03\4&2b4787d&0&0020" .
    (i) I/O should not be impacted.
    Error: 0x0, Error 0x00000000
        (x) Failover I/O application encountered an error issuing I/O to file: \Device\MPIODisk0
        File=drivers\storage\tests\mpiotest\logotest\test\testmanager.cpp Line=3377
    (i) Begin Simultaneous Bouncing Test
    ==================================================

    In both of two following drivers, a test fails.
     - Emulex LightPulse HBA Storport Miniport Driver of the Windows 2016TP5 standard(3/24/2015 10.2.370.8009)
     - Latest Emulex LightPulse HBA Storport Miniport Driver(1/26/2016 11.0.247.0)


    However, when I execute Multipath I-O test (LOGO) of HCK with the same test machine in Windows 2012/2012R2,
    Link Bouncing and Simultaneous Bouncing Test does Pass.
    The log at that time was as follows.
    ==================================================
    (i) Begin Link Bouncing Test
    (i) Do failover on chosen path...
    (i) cmdLine is: devcon disable "@PCI\VEN_10DF&DEV_F100&SUBSYS_F10010DF&REV_03\4&3b3eb1fd&0&0020" .
    (i) Do failback on chosen path...
    (i) cmdLine is: devcon enable "@PCI\VEN_10DF&DEV_F100&SUBSYS_F10010DF&REV_03\4&3b3eb1fd&0&0020" .
    (i) I/O should not be impacted.
    (i) Begin Simultaneous Bouncing Test
    ==================================================

    On why, a test of the 3rd party DSM of the same machine environment,
     In Windows 2016 TP5 + HLK, Multipath I-O test (LOGO) fails,
     In Windows 2012/2012R2 + HCK, will Multipath I-O test (LOGO) succeed?

    Because it may be the problem of the procedure at the time of this test practice, but does not understand a cause from log and setting information on hand,
    Please contact me if you may understand it from the following information output in the log of the error.
    ==================================================
    Error: 0x37, The specified network resource or device is no longer available.
        (x) Failed to send I/O [mode = 0x00], GLE: 0x00000037.
        File=drivers\storage\tests\mpiotest\logotest\test\ioapplication.cpp Line=431
    Error: 0x0, Error 0x00000000
        (x) Normal I/O did not complete successfully.
        File=drivers\storage\tests\mpiotest\logotest\test\ioapplication.cpp Line=493
    Error: 0x0, Error 0x00000000
        (x) Failover I/O application encountered an error issuing I/O to file: \Device\MPIODisk0
        File=drivers\storage\tests\mpiotest\logotest\test\testmanager.cpp Line=3377
    ==================================================
    Monday, June 13, 2016 12:23 AM

Answers

  • It is for Windows 2016 GA and seems to be revised.

    Bug:PDORemovePeriod Default Vaule is 0

    https://windowsserver.uservoice.com/forums/295047-general-feedback/suggestions/15047769-bug-pdoremoveperiod-default-vaule-is-0

    >This was a bug that was found in TP5. It will be fixed for GA.

    I expect it. Thank you.

    • Marked as answer by hirosea Wednesday, July 6, 2016 11:48 PM
    Wednesday, July 6, 2016 11:48 PM

All replies

  • After continuing an investigation, about movement after devcon disable, it was revealed that there was following difference by a combination of OS/DSM.

    +---------+--------+-------------------------+------------------------------------+
    |   OS        |  DSM     |        Just after all            | Time until LUs aren't recognized    |
    |               |              |        paths disable.          | after all paths disable.                   |
    +---------+---------+-------------------------+----------------------------------+
    | 2016TP5 |Our DSM |   LUs aren't recognized.  |       Immediately after.                  |
    +              +--------+------------------------+----------------------------------+
    |                | MSDSM  |   LUs are recognized.      |           1 minute.                            |
    +---------+--------+--------------------------+----------------------------------+
    | 2012      |Our DSM |    LUs are recognized.      |   A little less than 20 seconds      |
    +              +--------+------------------------+----------------------------------+
    |               | MSDSM  |    LUs are recognized.      |   A little less than 20 seconds       |
    +---------+--------+--------------------------+----------------------------------+

    It is the following things that it is thought from the table mentioned above.
    (1) Was there change in movement after all paths disable from Windows 2016TP5?
    (2) Was change in the processing also between the MSDSM of Windows 2012 and the MSDSM of Windows 2016TP5?
    (3) There is no difference in processing of Our DSM of Windwos 2012 and Windows 2016TP5 (It's a same binary .), but was movement on the OS/API side change?

    Isn't the person who knows information on a change of specifications of pass disable around here here?

    Thank you.



    • Edited by hirosea Thursday, June 23, 2016 12:56 AM
    Thursday, June 23, 2016 12:14 AM
  • When after trying it in various ways, I set 20 to the following Registry keys, Multipath I-O Test (LOGO) of HLK came to do Pass.
          HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\{3rd party dsm}\Parameters\PDORemovePeriod

    If there was not the Registry key mentioned above, to 2012R2, a value of PDORemovePeriod of MPIO seemed to be just used.
          HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\mpio\Parameters\PDORemovePeriod
    (if there was not the PDORemovePeriod key to MPIO, 20 of the default was used.)


    However, if there was not the PDORemovePeriod key to 3rd party dsm, in 2016TP5,
     0 was used regardless of having PDORemovePeriod of MPIO or not and  Link Bouncing Test of Multipath I-O Test (LOGO) was failed.

    I do not understand the reason why movement about PDORemovePeriod is changed in Windows2016TP5.
    I do this movement after Windows2016RTM, or please tell me if there is one to know.
    (Is it degrade or specifications change in Windows2016?)

    Thanking you in advance.

    Wednesday, June 29, 2016 12:00 AM
  • When, in environment of Windows2016TP5+MSDSM, I execute  Multipath I-O Test (LOGO) of HLK,
    60 is set to a value of the following registry automatically.
      HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\msdsm\Parameters\PDORemovePeriod

    After Windows2016TP5 installation, the Registry value of the above is not set,
    There is a set impression only to pass a test of Multipath I-O Test (LOGO) of MSDSM.


    When the registry value isn't established manually in case of 3rd party dsm, I don't pass a test,
    so I don't make it satisfied.
          HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\{3rd party dsm}\Parameters\PDORemovePeriod

    Will the information to need setting of PDORemovePeriod at the time of HLK execution be shown?

    • Edited by hirosea Wednesday, June 29, 2016 3:13 AM
    Wednesday, June 29, 2016 2:55 AM
  • When I read the following descriptions,

    http://www.networksteve.com/forum/topic.php/Get-MPIOSetting_Outputs_Incorrect_Value_PathVerificationState_-/?TopicId=113044&Posts=7
    "If the DSM does not set those registry values, then MPIO will default to its own registry values at which point the cmdlet would be reporting accurate data."

    It is certain that a change of the movement matched in 2016TP5.


    "The documentation isn't very good in this area and the cmdlet output is incomplete."
    I want you to make the right specifications / movement as the OS of Windows2016 clear.

    Thursday, June 30, 2016 1:53 AM
  • In PDORemovePeriod of Windows 2008, there seemed to be a similar bug.
    https://support.microsoft.com/en-us/kb/970525
    Monday, July 4, 2016 12:05 AM
  • It is for Windows 2016 GA and seems to be revised.

    Bug:PDORemovePeriod Default Vaule is 0

    https://windowsserver.uservoice.com/forums/295047-general-feedback/suggestions/15047769-bug-pdoremoveperiod-default-vaule-is-0

    >This was a bug that was found in TP5. It will be fixed for GA.

    I expect it. Thank you.

    • Marked as answer by hirosea Wednesday, July 6, 2016 11:48 PM
    Wednesday, July 6, 2016 11:48 PM