Forum Discussion

Genericus's avatar
Genericus
Moderator
13 years ago

Media server drive sharing - question about sharing and versions

I write all my backups to two virtual tape libraries for redundancy. (VTLA & VTLB)

I duplicate to physical tape using two mediaservers. (M3 and M4, each has two hba connections to each VTL)

I have 13 LTO5 physical drives which are shared on M3 and M4

All backups are under SLP control.

As the jobs complete and go to duplicate, I can get hundreds queued up as SLP determines the best combination of source and destinations.

 

I am finding that NetBackup will give control of a shared drive to a specific media server, and once it has control it will not give it up as long as it has jobs, so every now and then all the jobs will be on M3 or M4, which saturates the hbas from the VTL.

 

I had tried setting the drives so that some were only owned by M3 or M4, and others were shared, but found that there was no logic within NetBackup to use local drives first.

Here is my question - has this changed in any versions of netbackup? I think I struggled with this in 6.X, I am at 7.0.1 now and planning to upgrade - if this is improved in a version I would consider choosing that version.

6 Replies

  • I am not sure I understand the issue 100%. But I you wan't to limit the number of drives the SLP consumes the solution is straight for. Just create a new storage unit, and  limit the "maximum concurrent drives".

    Let's sat you have a storage unit called LTO5 with maximum concurrent drives to 13. Then create a new storage unit and call it LTO5-duplicate. Set "maximum concurrent drives to e.g. 4 and point the SLP's to use LTO-duplicate. Since there is no one to one relation between storage unit and physical drives, with works quite good.

  • Have you configured Storage Unit Groups?

    If so, what is Failover Policy set to?

  • I have storage groups defined for these physical drives, but only a single one for each of these media servers.

    Can I set up a primary local storage group and a secondary shared one?

    Seems like an intelligent solution if it would work.

     

    If I have 10 LTO5 drives, all in group M4-LTO5

    I can define 5 as M4-group1 and 5 as M4-group2 and have it use group1 first, then group 2.

    BUT - how do I specify which ones are in which group?

    As far as I know, I cannot.

  • The other issue is that sometimes I want it to have all drives on one media server. If all remaining duplications are on M4, it should have all the drives.

    However, once M4 gets all the drives, if M3 gets duplications queued up, it won't balance out until M4 clears up it's jobs.

     

  • Number of ways to handle this

    You can do groups as you say and add them all in or you can create copies of your actual storage units and reduce the number of drive for it - and then put those in a group if you wish

    So you have 10 drives you may want to create new storage units for each media server with say 7 drives for each rather then 10 and then put these in the group

    That way at least one of the media servers will get at least 3 drives to do its duplications - if you go down to 5 you risk wasting half of your drives when only one server has lots to do

    Do be careful with groups as you can find the Media servers passing data across the network for the other server to write to tape - you may find it better just to reduce the number of drives for each Storage Unit to allow the other media server to get a look in

    Hope this helps

  • In the STUG config what is rhe Selection Criteria/Failover Policy set to?

    Choices (Extract from Admin Guide I):

    ■ Prioritized. Choose the first storage unit in the list that is not busy, down, or out of media.
    ■ Failover. Choose the first storage unit in the list that is not down or out of media.
    If the Failover option is selected, when a job must queue for a storage unit, the job queues rather than try another storage unit in the group.
    ■ Round Robin. Choose the least recently selected storage unit in the list.
    If a storage unit is unavailable, NetBackup examines the next storage unit until it finds one that is  available.
    ■ Media server load balancing. Choose a storage unit based on a capacity-managed approach.
    If a storage unit is unavailable, NetBackup examines the next storage unit until it finds one that is available.

    A queue can form for a storage unit if the storage unit is unavailable.
    The following are some reasons why a storage unit can be considered unavailable:
    ■ The storage unit is busy.
    ■ The storage unit is down.
    ■ The storage unit is out of media.
    ■ The storage unit has no available space.
    ■ The storage unit has reached the Maximum concurrent jobs setting.