Forum Discussion

kumaran21's avatar
kumaran21
10 years ago
Solved

Extremely Slow Backup on Linux Client

How can I troubleshoot network issue with Netbackup and Linux clients.

I have Netbackup 7.1 running on Windows Server 2008 backing up several Linux and Windows clients.

In one Linux client in particular the backup runs extremely slow. 5.5 TB data takes more than 60 hours to complete. Recently had error 41 and restarted the netbackup client services. Now even the daily backups with 16GB data takes 18 hours to complete.

In comparision another Linux client with 7.7 TB data takes only 22 hours to complete.

Both the clients are connected to the network through bonded 10G connection to the same switch and the Netbackup LTO5 drives are connected to the same switch through a 10G fiber channel.

Please advice how I can rectify this issue.

Thanks in advance.

7.1.x and Earlier
Backup and Recovery
NetBackup
Windows Server (2003-2008)
  • Michal_Mikulik1's avatar
    Michal_Mikulik1
    10 years ago

    Hello,

    perform file copy test outside NetBackup - copy a file between this client and Media Server if it is also slow.

    If it won't be slow, then it could be a client with many small files, is not this the case?

    If it is, consider FlashBackup type of backup (see Snapshot Client Guide if it is supported for your client), or try splitting data volume into smaller paralelly backed-up portions - multistreaming on the client with multiplexing on the Media Server.

    Michal

8 Replies

Sort By
Replies have been turned off for this discussion
  • Michal_Mikulik1's avatar
    Michal_Mikulik1
    Moderator
    10 years ago

    Hello,

    perform file copy test outside NetBackup - copy a file between this client and Media Server if it is also slow.

    If it won't be slow, then it could be a client with many small files, is not this the case?

    If it is, consider FlashBackup type of backup (see Snapshot Client Guide if it is supported for your client), or try splitting data volume into smaller paralelly backed-up portions - multistreaming on the client with multiplexing on the Media Server.

    Michal

  • WVT's avatar
    WVT
    Level 4
    10 years ago

    Was it ever fast?

    What is the nature of the data?

    why are the 10gbe nics bonded?

    What is the networking on the media server?

    How long has everything been hooked up exactly as it is now?

  • manatee's avatar
    manatee
    Level 6
    10 years ago

    fwiw, in a similar case here, we created another media server for the problem client.

  • Marianne's avatar
    Marianne
    Level 6
    10 years ago

    The first step is to test read speed from disk as per previous posts. 

    See : 
    Measuring disk performance with bpbkar 
    http://www.veritas.com/docs/000076349 

    On a Unix/Linux client you can also do a tar backup of the folder/filesystem to /dev/null and check how long it takes:
    time tar cBf /dev/null /folder

    When read performance from disk is good, test network performance - ftp +- 2GB file to media server and see how long it takes.

     

  • Nicolai's avatar
    Nicolai
    Moderator
    10 years ago

    Use the  GEN_DATA directive.

    This will random generate data in memory from the client, but will transmit data all they way to tape as was it real data. Doing this will tell if the problem is tranmission lines or the problem is located on the client file system. And by the way, you can restore the data as well. But restored data will not occupy disk space, it will dissolve in memory on the client :-)

    http://www.veritas.com/docs/000091135