varyonvg & varyoffvg taking LONG time - Aix

This is a discussion on varyonvg & varyoffvg taking LONG time - Aix ; Hello, Got 2 nodes in cluster; both have access to same EMC disks, there is VG defined (oravg); oravg is imported on both nodes, but on one node varyonvg/varyoffvg takes ~3 seconds, on second node it is almost 2 minutes. ...

+ Reply to Thread
Results 1 to 3 of 3

Thread: varyonvg & varyoffvg taking LONG time

  1. varyonvg & varyoffvg taking LONG time

    Hello,

    Got 2 nodes in cluster; both have access to same EMC disks,
    there is VG defined (oravg); oravg is imported on both
    nodes, but on one node varyonvg/varyoffvg takes ~3 seconds,
    on second node it is almost 2 minutes. Any clues what may
    be the reason?

    There are no any entries in errpt, no any related errors
    in /var/log/messages, truss returns same details on both
    nodes, no idea where to look next help.

    regards
    Tomasz


  2. Re: varyonvg & varyoffvg taking LONG time

    Tinasz,
    It sounds like you've got some disk locking going on. If you're
    absolutely sure that the varyoffvg is completely done on Node1 and
    it's taking >2 minutes to read the disks on node2.
    You can try turning off reservation locking on the hdiskpowers on both
    nodes, which tells the LUNs to not allow a node to take exclusive
    ownership of the LUN. You still can only write to the LUN from 1 node
    at a time due to AIX LVM limitations. This is a fairly common issue
    with Dual VIO and HACMP setups.

    Anyway, the command is:
    chdev -dev $Disk -a reserve_lock=no
    chdev -dev $Disk -a reserve_lock=no -perm
    -perm will set the LUN to unlock at the first available opportunity.

    Good luck!


    On Sep 10, 5:43 am, ser...@gdziestamna.il.pw.edu.pl wrote:
    > Hello,
    >
    > Got 2 nodes in cluster; both have access to same EMC disks,
    > there is VG defined (oravg); oravg is imported on both
    > nodes, but on one node varyonvg/varyoffvg takes ~3 seconds,
    > on second node it is almost 2 minutes. Any clues what may
    > be the reason?
    >
    > There are no any entries in errpt, no any related errors
    > in /var/log/messages, truss returns same details on both
    > nodes, no idea where to look next help.
    >
    > regards
    > Tomasz



  3. Re: varyonvg & varyoffvg taking LONG time

    On 10 Sep, 10:43, ser...@gdziestamna.il.pw.edu.pl wrote:
    > Hello,
    >
    > Got 2 nodes in cluster; both have access to same EMC disks,
    > there is VG defined (oravg); oravg is imported on both
    > nodes, but on one node varyonvg/varyoffvg takes ~3 seconds,
    > on second node it is almost 2 minutes. Any clues what may
    > be the reason?
    >
    > There are no any entries in errpt, no any related errors
    > in /var/log/messages, truss returns same details on both
    > nodes, no idea where to look next help.
    >
    > regards
    > Tomasz


    Tomasz,

    Check that the OS and device drivers are identical on both nodes.
    Same for your cluster software.
    Check your FC cards and make sure that they are configured the same as
    the "fast" node. (lsattr -El )

    Has the SAN changed? Do you have the same number of paths to disks as
    the "fast" node?
    Check the zoning on the HBA's for both AIX nodes. Check the port
    lights on back of node.

    HTH,
    Duncan

+ Reply to Thread