HANFS and HACMP - Aix

This is a discussion on HANFS and HACMP - Aix ; Guys I have one situation , where a two nodes cluster with cascading RG configuration is working fine ( no HANFS configuration...) The shared disk is on DS4800 and it contains a filesystem ( /oltlg1) , which has been exported ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: HANFS and HACMP

  1. HANFS and HACMP

    Guys

    I have one situation , where a two nodes cluster with cascading RG
    configuration is working fine ( no HANFS configuration...)
    The shared disk is on DS4800 and it contains a filesystem ( /oltlg1) ,
    which has been exported from NodeB to NodeA over NFS...

    Problem is that , when NodeB fails , san disk moves from NodeB to Node
    A along with IP address without any problem , however application
    could not be started on Node A, as by any mean ,it is still looking
    for NodeA:/oltlg1;even NFS client on nodeA comes in some sort of hang
    situation ( df command output hangs)...application vendor is now
    saying that application still looking for NFS mounted filesystem
    rather than locally available filesystem...

    Is there any way to configure NFS on nodeB to export using service ip
    address , rather than using nodename of NodeB? Will this approach
    solve this issue?

    OR Should i go straightforward for HANFS configuration rather than
    simple cascading RG configuration? As per my understanding , it means
    just putting NFS as part of RG? Will it be a big change in HA
    configuration ( beside RG configuration changes ). I would appreciate
    any document link etc...

    Please advice

    Regards


  2. Re: HANFS and HACMP

    Polani schrieb:
    > Guys
    >
    > I have one situation , where a two nodes cluster with cascading RG
    > configuration is working fine ( no HANFS configuration...)
    > The shared disk is on DS4800 and it contains a filesystem ( /oltlg1) ,
    > which has been exported from NodeB to NodeA over NFS...
    >
    > Problem is that , when NodeB fails , san disk moves from NodeB to Node
    > A along with IP address without any problem , however application
    > could not be started on Node A, as by any mean ,it is still looking
    > for NodeA:/oltlg1;even NFS client on nodeA comes in some sort of hang
    > situation ( df command output hangs)...application vendor is now
    > saying that application still looking for NFS mounted filesystem
    > rather than locally available filesystem...
    >
    > Is there any way to configure NFS on nodeB to export using service ip
    > address , rather than using nodename of NodeB? Will this approach
    > solve this issue?
    >
    > OR Should i go straightforward for HANFS configuration rather than
    > simple cascading RG configuration? As per my understanding , it means
    > just putting NFS as part of RG? Will it be a big change in HA
    > configuration ( beside RG configuration changes ). I would appreciate
    > any document link etc...
    >
    > Please advice
    >
    > Regards
    >


    Hi,

    what you need to configure is a so called HACMP "crossmount" (Rule: All
    HACMP internal NFS mounts and exports must be controlled by HACMP)

    Generally it works that way:

    - Create your filesyestem to mount to /oltlg1_x (instead of /oltlg1)
    - Change your RG running on NodeB (and having to perform the crossmount)

    => Change/Show All Resources and Attributes for a Custom Resource Group

    Filesystems/Directories to Export (NFSv2/3) [/oltlg1_x]
    Filesystems/Directories to NFS Mount [/oltlg1_x;/oltlg1]
    Network For NFS Mount []


    Don't forget to set:
    Filesystems mounted before IP configured true

    And create the following /usr/es/sbin/cluster/etc/exports entry:
    /oltlg1_x
    -sec=sys,rw,access=NodeA:NodeAp:NodeB:NodeBp,root=N odeA:NodeAp:NodeB:NodeBp

    where: NodeAp = Node A's persistent label
    NodeBp = Node B's persistent label

    After that, synchronize cluster (a RECONFIG will occur) and crossmount
    should be there and look like this:

    On Node A:
    =========
    NodeA:/oltlg1_x 1690304512 167566488 91% 141333 1% /oltlg1

    On Node B:
    ==========
    $ df | grep oltlg1
    /dev/oltlg1_lv 1690304512 167566488 91% 141333 1% /oltlg1_x
    NodeA:/oltlg1_x 1690304512 167566488 91% 141333 1% /oltlg1

    Hope that solves your issue.

    PS: Node A needs to be a participating node in resource group containing
    the /oltlg1_x filesystem for the crossmount.

    Regards,
    Uwe Auer

  3. Re: HANFS and HACMP

    Polani schrieb:
    > Guys
    >
    > I have one situation , where a two nodes cluster with cascading RG
    > configuration is working fine ( no HANFS configuration...)
    > The shared disk is on DS4800 and it contains a filesystem ( /oltlg1) ,
    > which has been exported from NodeB to NodeA over NFS...
    >
    > Problem is that , when NodeB fails , san disk moves from NodeB to Node
    > A along with IP address without any problem , however application
    > could not be started on Node A, as by any mean ,it is still looking
    > for NodeA:/oltlg1;even NFS client on nodeA comes in some sort of hang
    > situation ( df command output hangs)...application vendor is now
    > saying that application still looking for NFS mounted filesystem
    > rather than locally available filesystem...
    >
    > Is there any way to configure NFS on nodeB to export using service ip
    > address , rather than using nodename of NodeB? Will this approach
    > solve this issue?
    >
    > OR Should i go straightforward for HANFS configuration rather than
    > simple cascading RG configuration? As per my understanding , it means
    > just putting NFS as part of RG? Will it be a big change in HA
    > configuration ( beside RG configuration changes ). I would appreciate
    > any document link etc...
    >
    > Please advice
    >
    > Regards
    >

    Hi,

    what you need to configure is a so called HACMP "crossmount" (Rule: All
    HACMP internal NFS mounts and exports must be controlled by HACMP)

    Generally it works that way:

    - Create your filesyestem to mount to /oltlg1_x (instead of /oltlg1)
    - Change your RG running on NodeB (and having to perform the crossmount)

    => Change/Show All Resources and Attributes for a Custom Resource Group

    Filesystems/Directories to Export (NFSv2/3) [/oltlg1_x]
    Filesystems/Directories to NFS Mount [/oltlg1_x;/oltlg1]
    Network For NFS Mount []

    Don't forget to set:
    Filesystems mounted before IP configured true

    And create the following /usr/es/sbin/cluster/etc/exports entry:
    /oltlg1_x
    -sec=sys,rw,access=NodeA:NodeAp:NodeB:NodeBp,root=N odeA:NodeAp:NodeB:NodeBp

    where: NodeAp = Node A's persistent label
    NodeBp = Node B's persistent label

    After that, synchronize cluster (a RECONFIG will occur) and crossmount
    should be there and look like this:

    On Node A:
    =========
    NodeB:/oltlg1_x 1690304512 167566488 91% 141333 1% /oltlg1

    On Node B:
    ==========
    $ df | grep oltlg1
    /dev/oltlg1_lv 1690304512 167566488 91% 141333 1% /oltlg1_x
    NodeB:/oltlg1_x 1690304512 167566488 91% 141333 1% /oltlg1

    Hope that solves your issue.

    PS: Node A needs to be a participating node in resource group containing
    the /oltlg1_x filesystem for the crossmount.

    Regards,
    Uwe Auer

  4. Re: HANFS and HACMP

    On Oct 27, 8:48 am, Polani wrote:
    > Guys
    >
    > I have one situation , where a two nodes cluster with cascading RG
    > configuration is working fine ( no HANFS configuration...)
    > The shared disk is on DS4800 and it contains a filesystem ( /oltlg1) ,
    > which has been exported from NodeB to NodeA over NFS...
    >
    > Problem is that , when NodeB fails , san disk moves from NodeB to Node
    > A along with IP address without any problem , however application
    > could not be started on Node A, as by any mean ,it is still looking
    > for NodeA:/oltlg1;even NFS client on nodeA comes in some sort of hang
    > situation ( df command output hangs)...application vendor is now
    > saying that application still looking for NFS mounted filesystem
    > rather than locally available filesystem...
    >
    > Is there any way to configure NFS on nodeB to export using service ip
    > address , rather than using nodename of NodeB? Will this approach
    > solve this issue?
    >
    > OR Should i go straightforward for HANFS configuration rather than
    > simple cascading RG configuration? As per my understanding , it means
    > just putting NFS as part of RG? Will it be a big change in HA
    > configuration ( beside RG configuration changes ). I would appreciate
    > any document link etc...
    >
    > Please advice
    >
    > Regards



    What is anyway the reason for the NFS cross mount ? and what was you
    initial goal.
    Even that i am not very familiar with hacmp anmore you configuration
    looks a little bit funny for me.

    cheers
    Hajo



    In my understanding on a simple HACMP configuration :

+ Reply to Thread