UNSOLVED

idokatz

updated

16 years ago

I

idokatz

3 Posts

0

32136

September 27th, 2010 03:00

Best Practice

Hello All,

I'm new and not familair with Dell PowerValut and I have a little question:

This is the scenerio:

I got two Physical PowerEdge servers which I want them to share the same network storage/lun storage which I want to configure in my MD3000I storage.

How can I do it?

I tried to do it as non-clusterd group and clusterd group and I saw the the drivres on the servers but when I did a change on server X I didn't see it on server Y?

 

Can you please assist me?

 

Thanks a lot :)

Ido Katz

 

  • mrokkam

    154 Posts

    593

    0

    Posted September 27th, 2010 08:00

    In order to share the storage across both the servers, you can install a clustered OS on the servers.

    If you just want to have shared storage space, it might be easier to map the storage to a single server and then share it as an NFS/CIFS share to the second server (in effect, running a NAS server on the server that is connected to the PowerVault MD array).

  • Dev Mgr

    6 Operator

    •

    9308 Posts

    •

    3 Points

    593

    0

    Posted September 27th, 2010 08:00

    2 servers can only share SAN disk storage (any SAN from any vendor) if the servers have some kind of blocking system in place to prevent the 2 servers from overwriting eachothers changes. For Windows this commonly is clustering.

     

    What happens is the following:

    - server boots

    - it reads the partition table (and knows where which partition starts and ends)

    - it reads the file allocation table (and caches (more or less) this index (that keeps track of where the pieces of each file are located))

    - when it performs a write it updates the file allocation table, but other than that it doesn't bother to re-read the file allocation table

     

    This last step is where the problem lies if 2 non-clustered servers share a SAN-based disk; one server performs the write, but the other server doesn't know (and has no reason to re-read the file allocation table to find out that a change occurred). It'll even eventually cause data corruption as at one point the 2nd server will make it's own writes and write data to a spot that it assumes is unused, but the other server wrote data to, so the other server's data is now (partially) gone.

  • idokatz

    3 Posts

    593

    0

    Posted September 27th, 2010 09:00

    Hey,

     

    1st of all I want to thank you for the explaination.

    2nd of all now I have a strange problem, i'm running the cluster verfication tool (windows 2008 server 64Bit R2 ENT Edition), and i'm getting again and again an error on scsi-3 presistent reservation test.

     

    This is the excat error:

    Failed to Register PR key for cluster disk 3 from node sql1.ef.tech status 1

     

    Now, I found in goog that it mate be something with my Dell MD3000I but it ain't because in this test there are 3 storage drivers (600GB,200GB and 50GB)

    and the test failed only when he trying to register on the 50GB one.

    My Firmware version is: 07.35.31.60 and i'm using Modular Disk Storage Manager Version: 3.56.G6.37

    You can download and view the log here:

    Normal 0 false false false EN-US X-NONE HE MicrosoftInternetExplorer4 https://rcpt.yousendit.com/955773799/7810b188bf240383a0420bd18d31d788

    I will really appriciate if you can help me with this one :(

     

    Thanks a lot

    Ido Katz

     

     

  • Dev Mgr

    6 Operator

    •

    9308 Posts

    •

    3 Points

    593

    0

    Posted September 27th, 2010 13:00

    Did you install the Modular Disk Storage Manager (at least the host part) on both servers and reboot? This is needed for MPIO to properly work, which is probably a requirement for the validation check(s) to pass.

  • idokatz

    3 Posts

    593

    0

    Posted September 28th, 2010 02:00

    Hi,

    I did it and it work :)

    But i'm getting something else now :( :

    Cluster disk 0 is a Microsoft MPIO based disk
    Cluster disk 0 from node sql1.ef.tech has 1 usable path(s) to storage target
    At least two independent paths to storage target are recommended for each cluster disk
    Cluster disk 0 from node sql2.ef.tech has 1 usable path(s) to storage target
    At least two independent paths to storage target are recommended for each cluster disk
    Cluster disk 1 is a Microsoft MPIO based disk
    Cluster disk 1 from node sql1.ef.tech has 1 usable path(s) to storage target
    At least two independent paths to storage target are recommended for each cluster disk
    Cluster disk 1 from node sql2.ef.tech has 1 usable path(s) to storage target
    At least two independent paths to storage target are recommended for each cluster disk
    Cluster disk 2 is a Microsoft MPIO based disk
    Cluster disk 2 from node sql1.ef.tech has 1 usable path(s) to storage target
    At least two independent paths to storage target are recommended for each cluster disk
    Cluster disk 2 from node sql2.ef.tech has 1 usable path(s) to storage target
    At least two independent paths to storage target are recommended for each cluster disk
    Getting information of registered DSMs from node sql1.ef.tech
  • JOHNADCO

    2 Intern

    •

    847 Posts

    593

    0

    Posted September 28th, 2010 06:00

    So on the ESX side, if you go to configuration, storage adapters, iSCSI adapter do you see multiple targets per lun?    If not, it sounds like you don't really have your iSCSI network setup correctly.    Generally a min of two nics for iSCSI and then two switches (one for each subnet) is standard on the iSCSI network.