none
DPM 2007 and ISCSI SAN RRS feed

  • Question

  • Hi,

    Basic issue is that I have 2 servers connected to a Dell MD3000i ISCSI SAN. One runs DPM 2007 and another runs Hyper-V. If the DPM server is switched on the ISCSI disk performance on the Hyper-V server is terrible. If the DPM server is switched off then the performance is fine. I am unable to open the disk management snapin mmc on the DPM server (keeps coming up as not responding).

    I am running DPM 2007 on Server 2008 SP2 X64 with a Dell 2950 with 16GB of RAM, 2 * Quad Core CPU, and connected to a Dell MD3000i ISCSI SAN that is presenting the server with approximately 6TB of storage over 3 LUNS.

    I have one other host that shares the MD3000i SAN storage and that is running Hyper-V on Server 2008 SP2 X64 with a Dell 2950 with 32 GB of RAM, 2* Quad Core CPU, and connected to the MD3000i ISCSI SAN that is presenting the server with 300GB of storage on 1 LUN.

    The MD3000i should provide approximately 80mbps write and 120mbps read IO performance on the ISCSI disks through the hosts.

    If the DPM 2007 server is switched on and running I get less than 10mbps on sequential reads when I run IO testing (IOMeter) on the Hyper-V host machine (testing it's 300GB) presented LUN. This is with all other IO on the hyper-V host switched off. If I switch off the DPM server performance on the Hyper-V host is good and gets approx 80mb write / 120mb read as I would expect.

    I am unable to run IOMeter on the DPM host as it keeps coming up as not respongding (I believe this is due to the amount of dynamic disks created by DPM) nor am I able to open the disk managment snapin (again not resonding).

    I have a call open with Dell and have checked the MD3000i hardware and config and there are no errors or misconfig... it looks to be squarely on the DPM box that the issue is occuring.

    Could DPM really hammer that much IO to make the other SAN host unusable? Any ideas?

     


    I ain't getting on no plane - fool!!!!!!!!!!!!
    • Moved by Praveen D [MSFT] Monday, July 19, 2010 8:40 AM Moving to DPM Storage Forum (From:Data Protection Manager)
    Thursday, May 27, 2010 12:36 PM

Answers

  • Can you please try bringing DPM service down(net stop dpm) first to lower the load and bring perform or other IO tools working then you can bring the DPM service up. Another way is to measure the IO happening for the VHDs/Disks on which DPM server VM service is running. Can you please try one of the approaches and see if it helps. As I said earlier 80Mbps is something that DPM might end-up using when its backing up data. Having too many dynamic disks/volume shouldn't create the problem as it is the DPM server and the the OS is running fine when those volumes are accessed. Accessing dynamic disk/volume information should not result in generating lots of data over SAN. Hope this helps you debug the issue one step further.


    Thanks, Praveen D [MSFT] This posting is provided "AS IS" with no warranties, and confers no rights.
    Tuesday, June 15, 2010 12:33 PM

All replies

  • If you are protecting sufficiently large amount of data on the DPM server then its very much possible that DPM using all the SAN speed. To my knowledge 80mb write speed is not very high for DPM server create. The best thing is to use perfmon/any other tool and monitor disk IO happening on the DPM server. That should give you the idea about idea. Please check if this happening always when DPM server is running or only when DPM server is running the backup jobs to isolate the other loads in DPM server VM.


    Thanks, Praveen D [MSFT] This posting is provided "AS IS" with no warranties, and confers no rights.
    Thursday, June 10, 2010 10:32 AM
  • Hi, the problem is that I am unable to run any io tools such as iometer, perfmon, or even open disk management mmc (keep hanging with not responding). I believe this is due to the amount of dynamic disks that dpm has created and the os and io tools have difficulty scanning all the disks hence the tools always hanging. Any advice?
    • Edited by rEMOTE_eVENT Thursday, June 10, 2010 11:07 AM Spelling
    Thursday, June 10, 2010 11:04 AM
  • Can you please try bringing DPM service down(net stop dpm) first to lower the load and bring perform or other IO tools working then you can bring the DPM service up. Another way is to measure the IO happening for the VHDs/Disks on which DPM server VM service is running. Can you please try one of the approaches and see if it helps. As I said earlier 80Mbps is something that DPM might end-up using when its backing up data. Having too many dynamic disks/volume shouldn't create the problem as it is the DPM server and the the OS is running fine when those volumes are accessed. Accessing dynamic disk/volume information should not result in generating lots of data over SAN. Hope this helps you debug the issue one step further.


    Thanks, Praveen D [MSFT] This posting is provided "AS IS" with no warranties, and confers no rights.
    Tuesday, June 15, 2010 12:33 PM