none
Backup exh2010 takes too long using wbadmin RRS feed

  • Question

  • I'm using the following statement to perform a full backup of my monolithic Exchange 2010

    C:\Windows\System32\wbadmin.exe start backup -backupTarget:\\CRMBUFFALO\BackupSpot_OffSite\MSExchange -include:C:,F: -allCritical -systemState -quiet

    It is a virtual machine 4 cores, 4Gb ram (2008 R2 + Exchange 2010) running on hyperv 2008 R2

    C drive is 30gb. databases on the F drive are about 350gb. 80 mailboxes.

    Note: The target folder is always created before starting the backup, every time.

    The backup needs 24h!

    Since all involved machine are on, at-least, 1gb LAN it really seems too much.

    The log does not show any error.

    Any hint about which could be the issue? Thank in advance. pf

    Monday, February 18, 2013 11:45 AM

Answers

  • I redo the backup and take a look around.

    It is a VM all i/o becomes network i/o at physical level. And I mean both HD and network since them all are virtualized

    I made some file trasfer, a couple of GB, for test and got good transfer rate, always.

    The exchange DBs reside onto one Clustered shared volume (CSV)

    When backup starts the CSV goes to "Redirected Access" mode. I feel that's the guilty.

    ok, I'm gonna work on that ra mode now!

    ciao

    p.s. yes,on the same destination seems to work quicker; from 24h to 14h...

    • Proposed as answer by Zi Feng Tuesday, February 26, 2013 2:30 AM
    • Marked as answer by Zi Feng Friday, March 1, 2013 1:44 AM
    Monday, February 25, 2013 10:18 AM

All replies

  • Your system seems short on RAM (not that this is why backups are taking a long time but needed to point that out). What speed disk do you have on the server and what speed of disk are on the backup location?  Also what is the network utilization look like for the NIC on the Exchange server?

    Jason Apt, Microsoft Certified Master | Exchange 2010 My Blog

    Tuesday, February 19, 2013 3:34 AM
  • Can you simple test by taking a backup on same server and then compare ? , look for CPU and RAM consumtion during backup
    Tuesday, February 19, 2013 5:01 AM
  • Sorry, ram mistaken. 8Gb...

    - The server is a VM. The F drive, with all dags, is a fixed-size vhd throu' a virtual scsi controller

    The below hw is a full HP Proliant Blade plus HP EVA 4400 with plenty of 15k 300gb scsi drives

    The virtual network adapter is the new Synthetic one.

    - The lan is 1gbits

    - The backup nas is a brand new Buffalo Terastation proquad; 4 1Tb sata drives.

    I'm gonna test it all next Saturday.

    One side question: if I repeated the backup on the same folder (with the previous copy), could it be quicker? In other word the backup task copies all data or is it smart enough to apply such a kind of "differential image"?

    Thank for any hint and sorry my English

    @Yogesh Kumar Malhotra, "on the same server" do you mean attach another vhd to the vm (I do not have space on present drives)?

    pf



    • Edited by pgfiore Tuesday, February 19, 2013 3:22 PM
    Tuesday, February 19, 2013 1:55 PM
  • Yes you can use another vhd on .

    Also, if you will take backup at same destination ideally it should be quicker but not sure , let me check it if I can find something.

    Tuesday, February 19, 2013 2:01 PM
  • The throughput, indicated by Windows drag&drop, is at least 30MByte/s. It is a tiny % of the 10Gbits synthetic adapter of course. At present the server is working and consequently ram is fully used and cpu is swinging between 20-50% [w3wp.exe tasks, used mainly by iphones, should be moved sooner or later btw :(]

    The Buffalo san could reach the 60Mbyte/s when used by a physical machine. However I feel it is not the lan/hd throughput the trouble...

    Tuesday, February 19, 2013 3:34 PM
  • I redo the backup and take a look around.

    It is a VM all i/o becomes network i/o at physical level. And I mean both HD and network since them all are virtualized

    I made some file trasfer, a couple of GB, for test and got good transfer rate, always.

    The exchange DBs reside onto one Clustered shared volume (CSV)

    When backup starts the CSV goes to "Redirected Access" mode. I feel that's the guilty.

    ok, I'm gonna work on that ra mode now!

    ciao

    p.s. yes,on the same destination seems to work quicker; from 24h to 14h...

    • Proposed as answer by Zi Feng Tuesday, February 26, 2013 2:30 AM
    • Marked as answer by Zi Feng Friday, March 1, 2013 1:44 AM
    Monday, February 25, 2013 10:18 AM
  • 400Gb on 12h with 4 core / 4Gb RAM seem good. You can try some perfmon during backup to check where it hurt the most (ram, cpu, network, disk), but I bet you'll get faster with more core.


    • Edited by Bruce JDC Monday, February 25, 2013 10:26 AM
    Monday, February 25, 2013 10:25 AM
  • 400Gb on 12h with 4 core / 4Gb RAM seem good. You can try some perfmon during backup to check where it hurt the most (ram, cpu, network, disk), but I bet you'll get faster with more core.


    4 cpu is the max for a hyperv VM. The VM is 10Gb too (not 4)

    Accept the bet. Also because I do not see the reason why cpu could impact on backup elapsed time so heavely...

    Monday, February 25, 2013 11:08 AM