Forum Discussion

Ajith_Sankar's avatar
4 years ago

Linux Filesystem Backups with ALL_LOCAL_DRIVES

I want to cover the maximum file/folders available on a Redhat Linux 7 server. Which one should i use from below selection

/ with cross mount point enabled or ALL_LOCAL_DRIVES without cross mount.

The server having, Many filesystem outside the root(/). Having multiple LVMs and in Veritas cluster

No NFS mounts.

When tried backup with both option, found / with cross mount point have 600MB backup size more than the other one.

Please suggest which one is the best or both are having  same effect ?

  • Marianne's avatar
    Marianne
    4 years ago

    With ALL_LOCAL_DRIVES, you can select 'Allow Multiple data streams' in Policy Attributes.
    A separate backup stream will be created for each mount point.
    This way, when one stream fails, troubleshooting is easier and you can rerun the failed stream only.

5 Replies

Replies have been turned off for this discussion
  • In my env I do this:

    ALL_LOCAL_DRIVES without cross mount points

  • Both options are valid.

    Using / with cross mount point, will be performed as a single stream backup, nice for tape based backup. Don't work really well for very large backups.

    ALL_LOCAL_DRIVES will create many concurrent streams, really idea for disk based backup, but may require more resource from the OS (and if done really dump, will tip over the OS at backup start time).

     

  • For a server that is in a cluster, I won't do either.

    When backing up the local hostnames, I would explicitly specify the non-shared mountpoints.

    For clustered data, I would configure a policy for the virtual hostname with appropriate policy type to backup shared data.
    e.g. - if Oracle is being clustered, configure Oracle policy to backup clustered instance.

    If you backup a database with regular Standard policy, the backup will not be any good, because database files are in inconsistent state. Restoring the backed-up files will result in Oracle (or any other database type) failing to start.

    Even if only file-level backups are needed, I would still configure a separate policy for the virtual hostname and shared mount points in Backup Selection. This way you are sure that the backup will follow the virtual hostname in case of a failover.

     

    • Ajith_Sankar's avatar
      Ajith_Sankar
      Level 6

      Thanks for the resposne. This is basically an application server in cluster and app team want everything to be covered on both the nodes include cluster drives and os files. I will take backup of both the nodes with standard ALL_LOCAL_DRIVES in one policy so it should cover OS files from both nodes and cluster drives from active node.

      • Marianne's avatar
        Marianne
        Level 6

        With ALL_LOCAL_DRIVES, you can select 'Allow Multiple data streams' in Policy Attributes.
        A separate backup stream will be created for each mount point.
        This way, when one stream fails, troubleshooting is easier and you can rerun the failed stream only.