Visit Open-E website
Results 1 to 5 of 5

Thread: ISCSI Data Replicaton + Vmware

  1. #1

    Default ISCSI Data Replicaton + Vmware

    Is anybody using the ISCSI replication between 2 DSS servers, with ESX 3.01 ?

    We currently have running :-

    2 DSS servers (Dell 2900)'s with 900 Ggig of storage on each for VMDK's (VMFS)

    1 DSS server (Supermicro), that will eventually be used to be the destination for the ISCSI replication (ONCE replication from multiple sources to a single destination is supported)

    2 Dell 6850's running ESX 3.01, running the VM's. ESX servers boot from DAS, VM's boot from ISCSI on the DSS boxes.

    This is currently working in production, but i was wondering if anyone has statred to use the ISCSI replication, with VMware and is it working for them. IE failover, having both the source and destination Luns visible in ESX ??

    Thanks all

  2. #2

    Default Update on ISCSI Volume replication with vmware

    Thought i would do an update on this.

    I have now set up the test environment and been doing some testing. Initally i had a few problems, during failover but these were soon rectified.

    Test setup

    2 HP DL380's one as the source the other as the destination
    Created 2 small ISCSI LUNS'to keep replication times down a bit. Luns are 30 gig on each server.
    REplication via direct gig connection between the servers
    ISCSI Lan over a seperate gigabit switch

    Vmware server

    HP DL380, with VMware ESX 3.0.2
    Using Hardware ISCSI HBA. Note that they are QLA4010 (not officially supported by Vmware anymore, but still work)

    TEST

    Created the replication task on the source server and allowed everything to sync up, took about 10mins for the task to get to a consistant state.

    Added the Source Lun into vmware (using ESX 3.0.2)
    Created the VMFS partition, and then installed 2 test vm's (windows 2003)

    1st test Clean swap over
    Shut down the 2 Virtual servers
    Stopped the replication task between the 2 DSS servers
    removed the Source LUN from the 1st DSS server

    Changed the 2nd DSS server's LUN to be a source from destination

    On the vmware side ran vmkfstools -V , checked in the VIC, that i could see the replicated LUN, and it was all there. Powered up the VM's again... All working fine.

    There were some small bits that needed doing to get to this stage but nothing too complex.

    Next test will be a total failue of the DSS server (Power the source off with VM's still running). I will post my findings soon

  3. #3

    Default

    Paul,

    A few questions if I may:

    1. Does this work over FC HBA's too?

    2. I didn't think open-e supported HA/fail-over yet, so I assume this would be a cold swap of the storage to get this to work, whereas a SANMelody solution is a true active/active HA failover solution (for a hefty cost no less).

    3. Is there a way to reach you via email to talk in more detail about your findings and configs?

    Thanks,
    Tom

    Quote Originally Posted by paul2kl
    Thought i would do an update on this.

    I have now set up the test environment and been doing some testing. Initally i had a few problems, during failover but these were soon rectified.

    Test setup

    2 HP DL380's one as the source the other as the destination
    Created 2 small ISCSI LUNS'to keep replication times down a bit. Luns are 30 gig on each server.
    REplication via direct gig connection between the servers
    ISCSI Lan over a seperate gigabit switch

    Vmware server

    HP DL380, with VMware ESX 3.0.2
    Using Hardware ISCSI HBA. Note that they are QLA4010 (not officially supported by Vmware anymore, but still work)

    TEST

    Created the replication task on the source server and allowed everything to sync up, took about 10mins for the task to get to a consistant state.

    Added the Source Lun into vmware (using ESX 3.0.2)
    Created the VMFS partition, and then installed 2 test vm's (windows 2003)

    1st test Clean swap over
    Shut down the 2 Virtual servers
    Stopped the replication task between the 2 DSS servers
    removed the Source LUN from the 1st DSS server

    Changed the 2nd DSS server's LUN to be a source from destination

    On the vmware side ran vmkfstools -V , checked in the VIC, that i could see the replicated LUN, and it was all there. Powered up the VM's again... All working fine.

    There were some small bits that needed doing to get to this stage but nothing too complex.

    Next test will be a total failue of the DSS server (Power the source off with VM's still running). I will post my findings soon

  4. #4

    Default

    Tom to answer your questions :-

    1. FC HBA, Open E supports the Qlogic 23** series as targets, and there is a post from GHarding who has this working (only with 2gig cards at present, but it does not say if he is using it with vmware or not. In the test lab i only had a couple of Fibre QLA2310's, i was able to create a FC LUN on open E. ESX could see the LUN, add it as storage and format it VMFS. However when i tried to create a vm on it create the virtual disk got to 99% and then hung the ESX server . I am awaiting a new HBA to test with, as it could have been one of those at fault.

    2. You are right Open-E does not support HA failover yet, so it is a cold swap over. For the first test it was to see if it worked correctly with ESX. And as you can see from the results it does. So i can now shut down the VM's running off of 1 DSS server Turn off that DSS server (to upgrade maybe), goto the second Dss server, change the replicated volume to source and click apply. Goto the ESX server, Rescan the ISCSI HBA and it sees the LUN, Then power the VM's back on.. Nothing else to change or do.

    I guess that in time Open-E will work on this to get to a solution where this will be possible, as i am sure that many people are asking for HA functionality.. (Open-E guys are you working on a high availibility solution ???)

    3. Yes i have sent a PM to you with my email address.

    I have been testing what happens if the Source DSS server gets turned off whilst live VM's are running and then swapping over to the replicated data. So far this has been ok, with standard 2003 servers, will have to try SQL and exchange next.. The only obvious thing that happens is that when the VM's get powered it is as if it had crashed (asked why it was not shut down cleanly). Well that was to be expected, so i will continue testing.

    I will eventually get around to doing some benchmarking tests on it, but no point until i am happy it works 100% of the time.

    When i get another FC HBA i will try it with Fibre. Only downside currently on the fibre side is that the replication still has to go over a normal NIC !!.

    Paul

  5. #5

    Default

    Paul,

    Do you feel confident that this is operating solidly enough for production? I tried this as well, but I selected a destination after I had created the vmfs volume. It failed for me, but I think I should of had replication working BEFORE I created the volume. I am going to use DSS with this instead of iSCSI Enterprise. Any gotchas I should be aware of?

    Thanks - roxer

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •