Forum Discussion

hraju's avatar
hraju
Level 4
15 years ago

Global Cluster, can you verify my config - setup - architecture ?

I need to build a Disaster Recovery using Veritas Cluster and VVR. I need to have HA on Oracle 10g and Tomcat. This should be cheap as possible. This is not so data-intensive application.
This would be only 2 servers setup. I will have 2 clusters. Each with 1 server. I will use local disks. Then there will be replication between and having a Global cluster.
This are the only 2 servers on network with no domain or DNS. So one of these servers should be DNS and Domain. I think the backup server should be domain and DNS server. Is this OKay ? Or the domain server should be on third server or both servers ? Isn't this (domain and DNS) a single point of failure ? I am not sure how to setup this.
I will use 5.1 VCS HA with VVR on Windows 2003 R2 SE. Any help or idea appreciated.


  • If you are concerned with loosing 1 site enirely then you should have domain controllers avaialable at each site.  Having doman controllers at each site is good if you loose communications completely between the sites.  The DC at both sites will ensure that site is usable even without access to anything from the other site.

    -Wally

8 Replies

  • Hi Hraju,

    Ideally, the domain features should be on nodes not in the cluster.  I'm not sure if Oracle has a problem running on domain controller but I have seen issues with MSDTC when used on domain controllers in a cluster.  I'm also not sure if you are going to use MSDTC in your configuration or not.

    You need to be aware that you may run into problems with using local storage.  If you have internal RAID controllers they may not allow the correct SCSI commands to the disks to allow the cluster to control them correctly.  I would suggest running the following test to see if you will have problems.

    1. Create  a clustered disk group with the internal disk(s)  (they have to different physical disks than what is used by the OS - system and boot disks.)
    2.  Deport the clustered disk group.
    3.  Import the clustered disk group.

    If you can create it, deport and import it again without errors then you will be fine.  If not you will need to contact Technical Support to see if there is something that they can do for you. 

    If you just have a problem creating the disk group, you may simply need to run the following command: 

       vxclus usesystembus on

    This will allow you to create clustered disk groups on the same bus that the system and boot disks are on.

    You might also be able to get around the use of a Domain all together.  The cluster can be configured manually without the wizards or domain.  However, you should contact your sale rep to talk to them about a support exception to maintain a supported configuration without a domain and to get the steps needed to configure it.

    Thanks,
    Wally

     

  • Thank you Wally for answer.

    I will test internall raid controller as you suggested.  I will not use MSDTC. But I am still not sure, if I should use domain controller on both Global cluster sites or only on one.
  • If you are concerned with loosing 1 site enirely then you should have domain controllers avaialable at each site.  Having doman controllers at each site is good if you loose communications completely between the sites.  The DC at both sites will ensure that site is usable even without access to anything from the other site.

    -Wally
  • Wally,

    So as you said, my configuration is not working when I create clustered Disk Group. When I deport and import it as cluster DG. It is failing. I already created support case.
    I have IBM x3650 with eight internal SAS disks. I created 3 RAIDed disks in bios:
    1) 2disks mirrored
    2) 4disks in RAID5
    3) 2disks mirrored

    Do you think that I should not use RAID at all ?

    I use Veritas Storage Foundation HA 5.1 with latest upgrade with "vxclus usesystembus on".
    I have one primary Disk Group with boot disk. Then I want to create clustered Disk Group with rest of the disks.
    Creating Clustered DG raise this error:
    V-76-58645-542 Unable to reserve a majority of dynamic disks group members. Failed to start SCSI reservation thread.
    Importing DG as clustered raise this error:
    V-76-58645-585 Failed to reserve the majority of disks

    I tried "vxclus enable -gClusterDG -p", "vxassist rescan", "vxassist refresh", "vxassist resetbus" with no avail.

    C:\Documents and Settings\Administrator.MERKUR>vxdisk list
    Name             MediaName   Diskgroup         DiskStyle  Size(MB)   FreeSpace(M
    B)     Status      EnclosureID      P#C#T#L#  ProvisionType  ProvisionedSize Ser
    ialNumber         VDID
    Harddisk0          Disk1    BasicDG              MBR      285145     245148
    Imported         DISKS@S          P4C1T0L0 -              -
                   600605B00177FDF012A7C27C84290582
    Harddisk1          Disk1    ClusterDG            GPT      855315     855315
    Offline          DISKS@S          P4C1T1L0 -              -
                   600605B00177FDF012A7C27C8431F30A
    Harddisk2          Disk2    ClusterDG            GPT      285019     285019
    Offline          DISKS@S          P4C1T2L0 -              -
                   600605B00177FDF012A7C27E8443914D

    C:\Documents and Settings\Administrator.MERKUR>vxdisk diskinfo Harddisk0
    Disk information
        Device Name :   Harddisk0
        Media Name  :   Disk1
        Disk Group  :   BasicDG
        Disk Style  :   MBR
        Length      :   298997121024
        FreeSpace   :   257056418304
        BusType     :   0
        EnclosureId :   DISKS@S
        Port        :   4
        Target      :   0
        Channel     :   1
        LUN         :   0
        Signature   :   2285879348
        Status      :   Imported
        Serial Number:
        Vendor ID   :   IBM
        Product ID  :   ServeRAID-MR10i
        Adapter     :   RAID Controller
        Comment     :
        Spare       :   No
        Reserved    :   No
        Hot Use     :   Yes
        Write Cache Enabled      :   No
        SCSI-3 Reserved          :   No
        Track Alignment
        VID/PID                  :    Default/Default
        Track Alignment Offset   :    64
        Track Alignment Setting  :    Disable
        Thin Provisioning
        Provision Type           :    -
        Provisioned Size         :    -
        Allocation Unit          :    -
        Veritas Disk ID (VDID)   :    600605B00177FDF012A7C27C84290582
        Subdisks
           Disk1-01

  • Hi hraju,

    Sorry for not getting back to you sooner.  I'm not sure if you are still having this issue or not.

    Using internal disks with a RAID controller can be problematic as some RAID controllers do not fully support all SCSI commands (like reservations) to the physical disks.  I'm guessing that your RAID controller is one that does not allow SCSI reservation commands and thus causing problems with importing the clustered disk group.

    You can do basically two items to resolve this.

    1.  Replace the RAID controller with a RAID controller or SCSI controller that does support SCSI-2 reservations.
    2.  Contact Technical Support to see if a patch is available for the version of the product that you are running to by-pass the SCSI-2 reservation requirement.  I think there maybe a patch available but it might be limited as to which versions of the product it is for.


    Thanks,
    Wally
  • Hello,

    Thank you for reply. I needed it asap, so I decided to make my solution this way (if youre interested):
    I created cluster from two servers with internal disks. Then I created new service group from Oracle template. I deleted MountV and VMDg (I dont mount volume) and added one script, that take care of replication migration/takeover (depends on what is possible). This is how the script look like:

    set ERRORLEVEL=
    echo date: %date% >>C:\script\log.log
    echo time: %time% >>C:\script\log.log
    :: check if P is primary RVG - replication source
    vxrlink status Prlink >>C:\script\log.log
    if ERRORLEVEL 0 GOTO END
    ping 192.168.110.20 -n 1
    if ERRORLEVEL 0 GOTO MIGRATE
    if ERRORLEVEL 1 GOTO TAKEOVER

    :MIGRATE
    echo "I see secondary S. Trying to migrate....">>C:\script\log.log
    vxrds migrate OracleRVG 192.168.110.10 >>C:\script\log.log
    echo Return code for migration: %ERRORLEVEL% >>C:\script\log.log
    IF %ERRORLEVEL% NEQ 0 GOTO TAKEOVER
    GOTO END

    :TAKEOVER
    vxrds -autofb takeover OracleRVG >>C:\script\log.log
    echo Return code for replication takeover: %ERRORLEVEL% >>C:\script\log.log
    GOTO END

    :END
    echo > c:\script\marker.id



    I think that Veritas could support this kind of replicated clusters officially, having special resource to monitor replication and do migration or takeover. This way it is possible to use only two servers with internall disks and on long distance locations (they dont use same disk array). I am thinking to put this into Symantec Ideas. What do you think ?
  • Hey hraju,

    I had the same requirement as you are. two single node clusters at two different sites and replicate data using VVR (All internal Disks) and on top Global Cluster.  I got a patch from Symantec to bypass the SCSI requirement when using internal disks. But I am finally ended up creating a cluster with two nodes (instead of two single clusters and Global Cluster) and replicate data between them using VVR. Unnecessarily brought Global cluster option....

    Thanks,
    Sai

  • I also dont use Global cluster, but only two node cluster with replication. What is number of this patch ? Can I have it ?
    thank you