NFS mount on HACMP 2 node cluster

From: Jan Pomni (pocztex_at_COGITO.UNIV.GDA.PL)
Date: 12/13/04

  • Next message: Green, Simon: "Re: NFS mount on HACMP 2 node cluster" 
    Date:         Mon, 13 Dec 2004 19:25:25 +0100
To: aix-l@Princeton.EDU

    I have a two node cascading HACMP cluster (AIX 5.2 ML3, HACMP 4.5) with
    4 resource groups.

    One of the nodes is exporting /data directory and another one mounts it
    using NFS. In case first node dies, second takes all resource groups.
    First goes live again, second is doing NFS exporting, first is mounting
    as a NFS client. Everything works fine.

    In /etc/filesystems (on every node) I have a stanza for local mounts:
    /data:
            dev = /dev/data_lv
            vfs = jfs
            log = /dev/jfslog1
            mount = false
            check = false
            type = sharedfs
            options = rw
            account = false

    All the exports and NFS mounts are done by HACMP.

    Lately I needed to modify one of the NFS parameters
    (on the NFS client side). What I found
    in a docs
    HACMP 4.5 Admin guide:

    Default NFS Mount Options for HACMP
    The default options used by HACMP when performing NFS mounts are "soft,
    intr". If you want hardmounts or any other options on the NFS mounts, specify
    these through the AIX SMIT NFS panels for Add a File SystemforMounting
    (fastpath: smit mknfsmnt).When adding these filesystems, be sure to choose
    the filesystems option to the "MOUNT now, add entry to
    /etc/filesystems or both?" field, and take the default value of no to
    "/etc/filesystems entry will
    mount the directory on system RESTART". This will add the options you
    have chosen into the
    /etc/filesystems entry created.
    This entry is then read by the HACMP scripts to pick up any options you
    may have selected.

    So this stanza in /etc/filesystem (after smit mknfsmnt) would be:
    /data:
            dev = "/data"
            vfs = nfs
            nodename = node01-serv
            mount = false
            options = bg,hard,intr
            account = false

    Unfortunatelly as this /data mount point is already defined I can not
    add it here. So I have no idea how to pass additional parameters while
    mounting this NFS directory by HACMP. Any ideas?

    -jp

  • Next message: Green, Simon: "Re: NFS mount on HACMP 2 node cluster"

    Relevant Pages

    • How to manage shared persistent local caching (FS-Cache) with NFS?
      ... patches again and have NFS work with them. ... coherency is ignored for mounts that have ... Imagine the client sees ... file its own cache: ...
      (Linux-Kernel)
    • Re: Any hope of KDE 3.5 in F10? I want it too !
      ... FWIW...I found that this behavior stopped when I went to autofs ... Do you mean change 'nfs' to 'autofs'? ... after years of doing smb mounts I've only just learned how to do nfs ... I can change the mounts simply by changing LDAP and ...
      (Fedora)
    • Re: nfs crossmount ?
      ... (exported via NFS from node one) ... about this that you have to give HACMP control about the NFS crossmount. ... have a WARNING during cluster sync and verification, ... You should end up with the following mounts (numbers for sure are ...
      (comp.unix.aix)
    • Re: [PATCH] NFS: Replace null dentries that appear in readdirs list [try #2]
      ... This should now be left to the NFS client. ... which mounts can cross servers, but that uses more or less the same mechanism. ... What's wrong with having the kernel do it? ... Automounting and auto-expiration both. ...
      (Linux-Kernel)
    • Re: nfs crossmount ?
      ... Node 1 owns a VG called data-vg in which resides only a LV ... (exported via NFS from node one) ... yes this is possible and HACMP has a function for this. ... These setting make HACMP to perform the NFS crossmount automatically for you on startup of the cluster (and remove the mounts during stop of the cluster manager). ...
      (comp.unix.aix)