[HPADM] SUMMARY: Easing Impact of SAN Changes to HW/Device Paths

From: Dave T. (davidlt77_at_hotmail.com)
Date: 07/25/05

  • Next message: Sue Pellerito: "[HPADM] SOX Audit"
    To: hpux-admin@dutchworks.nl
    Date: Mon, 25 Jul 2005 09:45:59 -0400
    
    
    

    My original post is at the very bottom.

    I received several replies. Thanks go to:

    * Mike Whitton
    * Shyam Hazari
    * M. Fearer
    * Gary Paveza
    * Stuart Abramson

    I had already mostly written scripts to handle my vgexports and vgimports,
    but they weren't working for a simple reason: I wasn't using the "-s"
    option with vgexport, so my map files weren't recording VGID's. Once I
    realized this, I added it, and my scripts now seem to work fine. I have
    attached the following scripts:

    * vgsnapshot (creates map files and can export VG's, including unmounting
    filesystems and doing other things)
    * vgrecover (uses map files to import VG's)
    * load_balance (balances the load across controllers when the load has all
    been shoved to one controller)

    These are lengthy, detailed scripts. Use a "-h" option for information
    about vgsnapshot and vgrecover, which doesn't execute anything but the help
    feature. To learn more about load_balance, execute the script, and then ^c
    after the information page appears.

    I've written these scripts in a fashion where they should work pretty much
    in any HP-UX/SAN environment without any changes to them. However, these
    scripts are provided without any promises that they will be flawless in your
    environment, so exercise caution.

    The scripts are very LONG. This is so they can be pretty much a no-brainer
    for junior SA's to use.

    I'm also including a couple scripts that Mike Whitton and M. Fearer
    graciously sent to me.

    Following are the responses I received, in the order I received them:

    ==========
    From : Mike Whitton <mike@eng.wayne.edu>
    Sent : Wednesday, July 20, 2005 2:46 PM
    To : "Dave T." <davidlt77@hotmail.com>
    Subject : Re: [HPADM] Easing Impact of SAN Changes to HW/Device Paths

    Go to previous message | Go to next message | Delete | Inbox
    Attachment : disk.pl (< 0.01 MB)

    I wrote the attached script. After you attach and zone the new switch and
    reboot the server you can look up what the old disk controllers were in the
    LVMTAB and then enter them into this script. Once run, it will recreate all
    of the old disk ctd numbers with the new info. Make sure to manually delete
    the old ctd numbers prior to running...

    mike

    #!/opt/perl/bin/perl
    #
    #
    #
    #

    open(LS, "ls -al /dev/dsk/c21*|");

    # brokenpath is the new path after the controller has changed
    $brokenpath = "c21t";
    # use oldpath with the prior disks name in vgdisplay
    $oldpath = "c17t";

    while(<LS>)
            {
            #print $_ . "\n";
            @lsvalue = split;
            print $lsvalue[5] . " " . $lsvalue[9] . "\n";
            $lsvalue[9] =~ s/$brokenpath/$oldpath/;
            print $lsvalue[9] . "\n";
            print "mknod $lsvalue[9] b 31 $lsvalue[5]\n";
            system("mknod $lsvalue[9] b 31 $lsvalue[5]");
            }

    ==========

    From : Shyam Hazari <shazari@gmail.com>
    Reply-To : Shyam Hazari <shazari@gmail.com>
    Sent : Wednesday, July 20, 2005 2:59 PM
    To : "Dave T." <davidlt77@hotmail.com>
    Subject : Re: [HPADM] Easing Impact of SAN Changes to HW/Device Paths

    Go to previous message | Go to next message | Delete | Inbox

    If you move a server from one switch another, it would definitely
    change the devices. If that happenes, I would just rebuild the
    /etc/lvmtab using vgscan.

    cp /etc/lvmtab /etc/lvmtab.old
    mv /etc/lvmtab /etc/lvmtab.orig

    vgscan( becareful with vgscan)

    Mount the filesystems

    -Shyam

    ==========

    From : <mfearer@cox.net>
    Sent : Wednesday, July 20, 2005 2:59 PM
    To : "Dave T." <davidlt77@hotmail.com>
    Subject : Re: [HPADM] Easing Impact of SAN Changes to HW/Device Paths

    Go to previous message | Go to next message | Delete | Inbox

    Dave,

    Here is something that we use in our environment. With a little tweaking,
    perhaps you could use these:

    [root@zerg: action]# cat vgexport_zerg.ksh | more
    #!/usr/bin/sh
    #
    # terran:/var/adm/move/terran/action/vgexport_.ksh
    #
    # vgexport the local VGs from terran:
    #

    . /usr/usd/scripts/move/.moverc

    HOST=$( hostname )

    # Don't run this if Oracle is running

    ps -ef | grep ora_pmon | grep -v grep
    if (( $? == 0 ))
    then
            # Oracle is running, halt !
            print "Oracle is running. Don't vgexport anything!"
            exit
    fi

    # Ask SA if he wants to vgexport volumes on this server..
    read ANSWER?"Do you want to vgexport live VGs (yes/no): "
    typeset -l LOW_ANSWER=$ANSWER

    if [[ $LOW_ANSWER != "yes" ]]
    then
            print "You didn't enter 'yes'. Script will exit without vgexport."
            print "If you want to vgexport VGs, run script again and answer
    'yes'."
            exit 1
    fi

    # Answer was yes. vgexport

    print "You answered 'yes'. vgexports will run right now."

    bdfm.ksh # bdf before umountall
    umountall ; mount /stand
    bdfm.ksh # bdf after umountall

    strings /etc/lvmtab | grep /dev | grep -v -e dsk -e vg00 | while read VG
            do
                    vgchange -a n $VG
                    vgexport $VG
            done

    if [[ $LOW_ANSWER != "yes" ]]
    then
            print "You didn't enter 'yes'. Script will exit without vgexport."
            print "If you want to vgexport VGs, run script again and answer
    'yes'."
            exit 1
    fi

    # Answer was yes. vgexport

    print "You answered 'yes'. vgexports will run right now."

    bdfm.ksh # bdf before umountall
    umountall ; mount /stand
    bdfm.ksh # bdf after umountall

    strings /etc/lvmtab | grep /dev | grep -v -e dsk -e vg00 | while read VG
            do
                    vgchange -a n $VG
                    vgexport $VG
            done

    print "vgexport successfully run!"

    [root@zerg: action]# cat vgimport_zerg.ksh | more
    #!/usr/bin/sh
    #
    # zerg:/move.zerg/scripts/vgimport_zerg.ksh
    #
    # vgimport the local VGs to zerg:
    #
    # VG Minor #
    #

    HOST=$(hostname)
    . /usr/usd/scripts/move/.moverc

    function vgimpf {
            VG=$1
            typeset -Z2 MN=$2
            mkdir /dev/$VG
            mknod /dev/$VG/group c 64 0x${MN}0000
            vgimport -s -m $MOVEHOST/mapfiles/mapfile.$HOST.$VG $VG
            vgchange -a y $VG
            vgcfgbackup $VG
         }

    vgimpf vg01 01

    ==========

    From : Paveza, Gary <gary.paveza@AIG.COM>
    Sent : Wednesday, July 20, 2005 9:25 PM
    To : "'Dave T.'" <davidlt77@hotmail.com>
    Subject : RE: [HPADM] Easing Impact of SAN Changes to HW/Device Paths

    Go to previous message | Go to next message | Delete | Inbox

    I would suggest that you create export map files for all your disks:

    vgexport -m <mapfile> -s -p -v /dev/vg*

    This creates a file to be used in vgimport and more importantly puts the
    VGID on each disk in the volume group.

    If you have a problem that changes the device files, do the following:

    vgexport /dev/vg*

    mkdir /dev/vg*
    mknod /dev/vg*/group c 64 0xYY0000
    vgimport -m <mapfile> -s -v /dev/vg*

    ---------------------------------------------------------
    Gary Paveza, Jr.
    Senior Systems Administrator - HP-UX CSE
    (302) 252-4831 - phone

    ==========

    From : Stuart Abramson <stuarta46@yahoo.com>
    Sent : Sunday, July 24, 2005 3:18 PM
    To : "Dave T." <davidlt77@hotmail.com>
    Subject : Re: [HPADM] Easing Impact of SAN Changes to HW/Device Paths

    Go to previous message | Go to next message | Delete | Inbox
    Whenever we are going to make SAN changes we record the VG info of the VGs
    (minor number, PVs, map files, LVs, etc..). That way if the device file
    names change, then we can vgexport/vgimport and get them back.

    We also record the Veritas Netbackup tape information with Veritas Netbackup
    commands, (I forget their names now, tpautoconf, or someting like that.) to
    recover later.

    I can send them to you if you want them...

    ----Original Message Follows----
    From: "Dave T." <davidlt77@hotmail.com>
    To: hpux-admin@dutchworks.nl
    Subject: [HPADM] Easing Impact of SAN Changes to HW/Device Paths
    Date: Wed, 20 Jul 2005 14:38:14 -0400

    Environment:

    HP-UX 11i servers
    Brocade SAN switches
    VA7410 disk arrays
    SureStore tape libraries

    When changes are made to the ports on the SAN switches that route to the
    disk or tape devices, it changes the hardware and device paths to them in
    the OS. Whenever this happens, it has been a laborious process to recover
    everything. Has anyone found an easy way to handle this? Does anyone have
    scripts that ease the pain when the paths change?

    Dave

    --
                 ---> Please post QUESTIONS and SUMMARIES only!! <---
            To subscribe/unsubscribe to this list, contact 
    majordomo@dutchworks.nl
           Name: hpux-admin@dutchworks.nl     Owner: 
    owner-hpux-admin@dutchworks.nl
    Archives:  ftp.dutchworks.nl:/pub/digests/hpux-admin       (FTP, browse 
    only)
                http://www.dutchworks.nl/htbin/hpsysadmin   (Web, browse & 
    search)
    
    
    
    

    --
                 ---> Please post QUESTIONS and SUMMARIES only!! <---
            To subscribe/unsubscribe to this list, contact majordomo@dutchworks.nl
           Name: hpux-admin@dutchworks.nl     Owner: owner-hpux-admin@dutchworks.nl
     
     Archives:  ftp.dutchworks.nl:/pub/digests/hpux-admin       (FTP, browse only)
                http://www.dutchworks.nl/htbin/hpsysadmin   (Web, browse & search)
    



    • application/octet-stream attachment: vgrecover

  • Next message: Sue Pellerito: "[HPADM] SOX Audit"

    Relevant Pages

    • Conversation with a deaf person (Extracts) about the Phaistos Disk.
      ... scripts because knowing how to read onr artifact, ... B)- But the Proto-Ionian Theory can be easily integrated in the ... Disk ONLY because they know that studying it seriously would led to ...
      (sci.lang)
    • Re: Conversation with a deaf person (Extracts) about the Phaistos Disk.
      ... scripts because knowing how to read one artifact, ... B)- But the Proto-Ionian Theory can be easily integrated in the ... Disk ONLY because they know that studying it seriously would led them ...
      (sci.lang)
    • *nix data wipe tools
      ... The purpose here is to simplify regular maintenance. ... The scripts are meant to clean large disk areas safely and conveniently while ... The WipeFree scripts will securely wipe un-allocated disk space, ...
      (Focus-Linux)
    • *nix data wipe tools
      ... The purpose here is to simplify regular maintenance. ... The scripts are meant to clean large disk areas safely and conveniently while ... The WipeFree scripts will securely wipe un-allocated disk space, ...
      (Full-Disclosure)
    • *nix data wipe tools
      ... The purpose here is to simplify regular maintenance. ... The scripts are meant to clean large disk areas safely and conveniently while ... The WipeFree scripts will securely wipe un-allocated disk space, ...
      (Bugtraq)