Restoring SAN File System cluster configuration

This topic explains how to restore the configuration for the SAN File System cluster.

Steps

  1. If the system was backed up using the LUN method, and the entire cluster is down, perform these steps to restore the cluster configuration information:
    1. If you have previously saved the configuration files to another location, copy these files onto the boot drive for the engine.
      1. Copy Tank.Bootstrap to /usr/tank/server/config.
      2. Copy Tank.Config to /usr/tank/server/config.
      Note: If you have saved any other administrative configuration files, you can reference them when restoring the SAN File System metadata configuration.
    2. If the cluster bootstrap file, Tank.Bootstrap, is corrupted or missing, you can attempt to recreate the contents of that file using information from the metadata LUNs:
      1. Use the /usr/tank/server/bin/tank lsdisklabel –device command to find the master volume. If you cannot remember which device is your master volume, this is an iterative process of searching all suspected master volume devices until the command indicates you have found a valid master volume.
      2. Use the /usr/tank/server/bin/tank extractbootrecord command to regenerate Tank.Bootstrap from the master volume.
      3. Use the /usr/tank/server/bin/tank resetcluster command to reinitialize the master volume for subsequent rebuilding of the cluster configuration.
      4. Use the /usr/tank/server/bin/tank addserver command for all subordinate metadata server engines to recreate the cluster definition.
  2. If the system was backed up using the LUN method, and only the master metadata server is down, perform these steps to restore the cluster configuration information:
    1. Change the master metadata server to one of the subordinate metadata servers:
      1. Run /usr/tank/server/bin/tank lsserver to verify that the cluster does not have a master metadata server.
      2. Run /usr/tank/server/bin/tank lsengine to verify that the engine hosting the master metadata server is shut down.
      3. Run /usr/tank/server/bin/tank lsserver to verify that the metadata server to which you are going to assign the role of master is up and running.
      4. Access the subordinate metadata server that you want to set as the new master metadata server.
      5. Run the /usr/tank/server/bin/tank setmaster command to set it as the new master metadata server.
      6. If the previous master metadata server was managing any filesets, reassign those filesets to other metadata servers in the cluster.
    2. If you have previously saved the configuration files to another location, copy these files onto the boot drive for the engine.
      1. Copy Tank.Bootstrap to /usr/tank/server/config.
      2. Copy Tank.Config to /usr/tank/server/config.
      Note: If you have saved any other administrative configuration files, you can reference them when restoring the SAN File System metadata configuration.
    3. If the cluster bootstrap file, Tank.Bootstrap, is corrupted or missing, you can attempt to recreate the contents of that file using information from the metadata LUNs:
      1. Use the /usr/tank/server/bin/tank lsdisklabel –device command to find the master volume. If you cannot remember which device is your master volume, this is an iterative process of searching all suspected master volume devices until the command indicates you have found a valid master volume.
      2. Use the /usr/tank/server/bin/tank extractbootrecord command to regenerate Tank.Bootstrap from the master volume.
      3. Use the /usr/tank/server/bin/tank resetcluster command to reinitialize the master volume for subsequent rebuilding of the cluster configuration.
      4. Use the /usr/tank/server/bin/tank addserver command to add the original master metadata server to the cluster.
    4. Repeat steps 2.a.a through 2.a.f to change the master metadata server back to the original metadata server.
    5. Use the /usr/tank/server/bin/tank resetcluster command to reinitialize the master volume for subsequent rebuilding of the cluster configuration.
  3. If the system was backed up using the API method, perform these steps to restore the cluster configuration information:
    1. If you have previously saved the configuration files to another location, copy these files onto the boot drive for the engine.
      1. Copy Tank.Bootstrap to /usr/tank/server/config.
      2. Copy Tank.Config to /usr/tank/server/config.
      Note: If you have saved any other administrative configuration files, you can reference them when restoring the SAN File System metadata configuration.
    2. If you suspect that the metadata LUNs are corrupted, you can perform these steps to recreate the cluster definition:
      1. Delete all Tank.Bootstrap and Tank.Config files from your metadata server engines.
      2. Start the /usr/tank/server/bin/tank binary on your master metadata server with the install option rather than normal option.
        Attention: The existing metadata data server information will be overwritten.
        This will create new Tank.Bootstrap and Tank.Config files on your metadata server master. Be sure to specify the same cluster name that was used prior to the disaster:
      3. Now start the master metadata server with /usr/tank/server/bin/tank normal command.
      4. Use the addserver command to add all subordinate metadata server engines. This will create new Tank.Bootstrap and Tank.Config files on the subordinates.

Parent topic: Managing disaster recovery

Related reference
tank extractbootrecord
tank lsdisklabel
tank resetcluster

Terms of use | Feedback
(C) Copyright IBM Corporation 2003, 2004. All Rights Reserved.