What is the best way to connect our data volumes to a virtual server

[Aalbrecht Jobsen]

We have the following dilemma:

In our LeftHand 9.5 and VSphere 4.1 environment:

We can connect our SQL or Exchange DATA and log volumes as RAW devices, but than we are unable to make a application managed snap from the datasore (VMFS volume),

or we can connect our SQL or Exchange DATA and log volumes with iSCSI initiator from the Windows server, but than an application managed snapshot took about 25 minutes with all VM's frozen in that datastore,

or we can put the volumes in the datasore as virtual disks, but that is not advised and not so very practical for servers with big data volumes.

Question what is the best practice for application managed snapshot and is VSphere 4.1 and LeftHand 9.5 application managed snapshots from the VMFS luns supported?

Thanks for any reply or hint,

Aalbrecht Jobsen

HZ University of applied sciences

[Paul Hutchings]

We mount ours from in the guest.

Never encountered the issue you mention making application managed snapshots with Exchange 2010, but we're using SANiQ 9.0 (not that I would expect it to happen with any version of SANiQ).

[Bryan McMullan]

Same as Paul here.  Exchange and SQL data volumes are mounted within the guests using the iSCSI Initiator. 

How many volumes did you have snapshotting at the same time when you had the 25 minute freeze?  It's likely that you're doing too many at the exact same time which may have overloaded your SAN nodes.  You typically want to stagger snapshots to reduce the overall load on the SAN. 

The only time I've ever had a freeze such as what you are describing is when I mistakenly set snapshots to go at the same time which caused a loaded, then overloaded scenerio.  Staggering fixed my problem (but I wasn't locked for 25 minutes).  If you are already staggering, you may want to contact support as you may have an issue with one or more of your nodes.

[Aalbrecht Jobsen]

So, iSCSI is the best way connect the volumes to the VM's.

But we still we see the very long time to make a snapshot.

When we make a snapshot from a single server through the VShere Centre with Quiesce guest file system and Snapshot the virtual machine's memory option marked on then we have the same long time to make a snapshot.

So it seems that the application managed snapshot from the Lefthand also makes a snap from the virtual machine's memory. When VM has for instance 8 GB memory it took a few minutes to copy the memory data to the snapshot.

There is no reason to snap the memory for disaster purposes, only a quiesce quest file system snap.

Is there any configuration in the LeftHand schedule to make that only the right snap will be taken?

We use ESXi4.1 at this moment.

Aalbrecht Jobsen

HZ University of applied sciences

[Jitun]

Is your SQL or Exchange Volume presented to the Virtual Machine ?
Are they on the P4000 SAN ?

If yes then you need to install Application Aware Snapshot on those virtual Machines.. and take App managed Snapshots of those Volumes from within the OS.

If you are taking Snapshot from VCenter, then you are taking a Snapshot of the Virtual Machine.

From what I understand you want the Snapshot of the SQL and Exchange only ?

[ArjanA]

Hi Aalbrecht,

If the VM is using the iSCSI initiator to access the volume, you need to install the Application Aware Snapshot software from the download directory of the CMC.

Kind regards,

Arjan