Forum Discussion

M_Danish's avatar
M_Danish
Level 3
11 years ago

Solaris 10 3/05 (FCS) Sparc and SFHA 4.1 Upgrade

 

Hi All

Here are the details of my setup

OS = Solaris 10 3/05 - Initial Release [First Customer Shipment]

Arch = Sparc

SFHA = 4.1 [VxFS, VxVM, VCS]

IMPORTANT **************

VCS Type =  2 Node Failover Production Cluster.

 

I'm planning to upgrade my OS to Solaris 10 1/13 update 11 through Liveupgrade after that will Upgrade SFHA using below upgrade path


(Current version) SFHA 4.1 ---- > 4.1 MP2 ---- > 5.1SP1 ---- > 5.1SP1RP4


Now my doubt is after the upgrade of OS and SFHA on one node if i switch the service groups from non-upgraded node to upgraded node will the Service groups switch over because OS and SFHA will be of different version.

Reply is highly appreciated.

 

Thanks and Kind Regards,

Danish.

Best Practice
business continuity
Cluster Server
Compatibility
Configuring
Solaris
Tip-How to
Upgrade
  • Gaurav_S's avatar
    Gaurav_S
    11 years ago

    Hi Danish,

    Yes, you need to stop GAB, Fencing & LLT all 3 of them will be upgraded to latest version.

    After doing hastop -local (this will stop HAD)

    you need to stop fencing first. To do this,

    /etc/init.d/vxfen stop

    modinfo |grep -i vxfen   (if you find module loaded)

    modunload -i <module_id>

    Then stop GAB

    /etc/init.d/gab stop  (this is same as doing gabconfig -U)

    modinfo |grep -i gab   (if you find module loaded)

    modunload -i <module_id>

    /etc/init.d/llt stop  (this is same as doing lltconfig -U)

    modinfo |grep -i llt   (if you find module loaded)

    modunload -i <module_id>

    For Solaris 10, you can use svcadm commands to offline the particular service. However I would still recommend to keep checking modules using modinfo command & unload any module before you stop the next component. Your order of shutting down should be

    1.HAD

    2. Vxfen

    3. GAB

    4. LLT

     

    G

8 Replies

Sort By
  • M_Danish's avatar
    M_Danish
    Level 3
    11 years ago

    Hi,

    I have the below queries please

    1) In order to stop VxVM to load at system boot time, we need to modify /etc/system file.

       What entries are to be commented out ?

       Is it only

    rootdev:/pseudo/vxio@0:0
    set vxio:vol_rootdev_is_volume=1

       (OR) also below entries are to be commented out ?

    forceload: drv/vxdmp
    forceload: drv/vxio
    forceload: drv/vxspec

    2) Once vxfen, gab & llt modules are unloaded to upgrade to 4.1MP2, should i again unload these modules
        to further upgrade to 5.1SP1 and again to 5.1SP1RP4 (OR) 6.0 ? After each upgrade should i stop the
        services in /etc/init.d and unload modules (OR) stopping services & unloading modules only once is
        enough to further upgrade to other versions ?

        My Plan is to upgrade from 4.1---> 4.1 MP2---> 5.1SP1--->5.1SP1RP4 (OR) 6.0

    3) Before upgrading should i also stop & unload the below listed modules

       24  12800a8  26920 268   1  vxdmp (VxVM 4.1z: DMP Driver)
       25 7be00000 2115c8 269   1  vxio (VxVM 4.1z I/O driver)
       27  12a4698   13f0 270   1  vxspec (VxVM 4.1z control/status driver)
      213 7b2d7528    c40 272   1  vxportal (VxFS 4.1_REV-4.1B18_sol_GA_s10b)
      214 7ae00000 1706a8  20   1  vxfs (VxFS 4.1_REV-4.1B18_sol_GA_s10b)

      If yes, should i stop & unload after each upgrade (OR) doing it once is enough ?


    4) Once the OS comes up with native disks (c#t#d#s#), In order to bring it under VxVM control we need to 
        encapsulate using vxdiskadm. My doubt is will rootdg, rootvol, plexes & subdisks be created
        automatically? Need a little clarification regarding this please.



    Response is highly appreciated as always,
    Thank you very much.

    Regards,
    Danish.

     

     

     

     

  • Gaurav_S's avatar
    Gaurav_S
    Moderator
    11 years ago

    Hi Danish,

    Yes, you need to stop GAB, Fencing & LLT all 3 of them will be upgraded to latest version.

    After doing hastop -local (this will stop HAD)

    you need to stop fencing first. To do this,

    /etc/init.d/vxfen stop

    modinfo |grep -i vxfen   (if you find module loaded)

    modunload -i <module_id>

    Then stop GAB

    /etc/init.d/gab stop  (this is same as doing gabconfig -U)

    modinfo |grep -i gab   (if you find module loaded)

    modunload -i <module_id>

    /etc/init.d/llt stop  (this is same as doing lltconfig -U)

    modinfo |grep -i llt   (if you find module loaded)

    modunload -i <module_id>

    For Solaris 10, you can use svcadm commands to offline the particular service. However I would still recommend to keep checking modules using modinfo command & unload any module before you stop the next component. Your order of shutting down should be

    1.HAD

    2. Vxfen

    3. GAB

    4. LLT

     

    G

  • M_Danish's avatar
    M_Danish
    Level 3
    11 years ago

    Hi,

    Version of SFHA (VxVM,VxFS,VCS) in my environment is 4.1

    I have planned to upgrade it to 5.1SP1RP4 (OR) to 6.0


    Before proceding with upgradation i will stop HAD on the local host by #hastop -local. Howerver should i also stop GAB & LLT. If yes what is the process to stop GAB & LLT before upgrading.

    Should i issue below commands before upgrading ?

    # /sbin/gabconfig -U
    # /sbin/lltconfig -U

    (OR)

     # modunload -i gab_id
     # modunload -i llt_id


    bash-3.00# lltconfig
    LLT is running
     

    bash-3.00# gabconfig -a
    GAB Port Memberships
    ===============================================================
    Port a gen   e7cc06 membership 01
    Port h gen   e7cc0c membership 01


    bash-3.00# svcs -a | egrep -i "vcs|llt|gab"
    legacy_run     Mar_20   lrc:/etc/rc2_d/S70llt
    legacy_run     Mar_20   lrc:/etc/rc2_d/S92gab
    legacy_run     Mar_20   lrc:/etc/rc3_d/S99vcs
    bash-3.00#

     

    bash-3.00# modinfo | egrep -i "vcs|llt|gab|vxfs|vxvm|fenc*"
     24  12800a8  26920 268   1  vxdmp (VxVM 4.1z: DMP Driver)
     25 7be00000 2115c8 269   1  vxio (VxVM 4.1z I/O driver)
     27  12a4698   13f0 270   1  vxspec (VxVM 4.1z control/status driver)
    213 7b2d7528    c40 272   1  vxportal (VxFS 4.1_REV-4.1B18_sol_GA_s10b)
    214 7ae00000 1706a8  20   1  vxfs (VxFS 4.1_REV-4.1B18_sol_GA_s10b)
    218 7b3ac000  22470 275   1  llt (LLT 4.1)
    219 7af60000  46ac8 276   1  gab (GAB device 4.1)
    220  13886b8  39488 277   1  vxfen (VRTS Fence 4.1)
    bash-3.00#

     

     

    Thanks & Regards,
    Danish.

     

  • Marianne's avatar
    Marianne
    Level 6
    11 years ago

    How do i stop GAB & LLT services.

     

    # /sbin/gabconfig -U
    # /sbin/lltconfig -U
                
    Unload the GAB and LLT modules from the kernel.
    Determine the kernel module IDs:

      # modinfo | grep gab
      # modinfo | grep llt
                        
    Unload the module from the kernel:

      # modunload -i gab_id
      # modunload -i llt_id
                      

    how do i temporarily stop HAD,GAB & LLT to come online after each reboot.

    # /usr/sbin/svcadm disable llt
    # /usr/sbin/svcadm disable gab
    # /usr/sbin/svcadm disable vcs

     

    My understanding of 'The script cannot remove patches...' is that the script can only install patches, it cannot later be used again should you wish to remove any of the MP2 installed by the install_vp script at a later stage.

  • M_Danish's avatar
    M_Danish
    Level 3
    11 years ago

    Thanks Gaurav bhai/starflyfly/Venkat garu for your response.

    Need your response to below query also please..

    Before upgrading OS on passive node when i issue 'hastop -local' it stops the HAD daemon on the node, but GAB & LLT will continue to run. How do i stop GAB & LLT services.

    Also there will be several reboots while upgrading OS & SFHA so how do i temporarily stop HAD,GAB & LLT to come online after each reboot.

     

    In the below link it is mentioned that "The install_vp script checks for installed VERITAS packages and installs the appropriate patches. The script cannot remove patches, so you must use the patchrm command to remove individual patches".

    Here i'd like to know the removal of which patches is being referred to. Current version of SFHA is 4.1 and after upgrading to 4.1 MP2 , will all patches related to 4.1 be removed.


    http://www.symantec.com/business/support/index?page=content&id=TECH50972

     

    Thanks and Regards,
    Danish.

     

  • Venkata_Reddy_C's avatar
    Venkata_Reddy_C
    Level 4
    11 years ago

    Running two different versions on the cluster nodes is not a supported configuration. You may be risking your production by planning upgrade in this manner. 

     

    The recommended is to upgrade both the nodes same time and online your applications. There may be more downtime in this case but upgrade will be clean.

     

    Regards,

    Venkat

  • Gaurav_S's avatar
    Gaurav_S
    Moderator
    11 years ago

    Hi Danish,

    OS version is not a major problem (though recommended is to have same version) but for sure, different SF version can not form a cluster.

    So, if your node 1 is on 4.1 while other node is upgraded to 5.1SP1RP4, they can't join cluster together. You will need to offline all resources from node 1 & online them on node 2 manually. However be careful with types.cf file. With new versions new bundled agents & definitions are rolled out. You can find this types.cf file in /etc/VRTSvcs/conf directory of upgraded node. This file needs to be replaced with types.cf file under /etc/VRTSvcs/conf/config directory so that VCS can use it. Some major differences were with diskgroup agent & any agent which is zone aware.

    Once you replace types.cf on upgraded node (node 2), offline everything on node 1 & then trigger online of all service groups on node 2 (hopefully should come online). Once everything is online on node 2, go for upgrade on node 1.

    Suggestion : there are 2 major releases after 5.1, you can consider to upgrade to 6.1 rather than to 5.1 which is pretty old now.

     

    G