Minimum size of datastore on vsphere 4.1

[mahesh414]

I have problem creating 1GB datastore with 1GB LUN size.  The error is

"THe capacity value  must be at least 1.3GB. Enter another value.

Please help. I need it urgent

thanks all

[a_p_]

Welcome to the Community,

I don't think there's much you can do except for providing the minimum disk space. Although the minimum size is 1.3GB, the recommended minimum size is 2GB.

Is this a test system? How does your setup look like? Maybe there's a chance for a workaround in case this is a virtual test system.

André

[mahesh414]

Thanks for your reply,

In fact I did spoke to my SAN team to allocate 2GB LUN instead of 1GB. Now I am able to create 2GB datastore.

reason being we are spposed to use flat VMDK files instead of RDM's.

By this what is the impact on the performance?? will it improve or detoriate?

[a_p_]

reason being we are spposed to use flat VMDK files instead of RDM's.

To me this looks like it is the other way around!? With the size of 1GB I'd assume this disk is used as e.g. a cluster quorum.

By this what is the impact on the performance?? will it improve or detoriate?

Actually there should not be a major difference in performance anymore. If performance is required, make sure you create the virtual disks as EagerZeroedThick disks and have the OS partitions aligned properly. The latter is important for RDMs too.

André

[mahesh414]

yes you are right we are using 1GB LUN as quorum disk. can u tell me if this approach is correct?

Also plese tell me the diference between

• Thick Provision Lazy Zeroed
• Thick Provision Eager Zeroed

thank you

[a_p_]

yes you are right we are using 1GB LUN as quorum disk. can u tell me if this approach is correct?

It depends on your MSCS setup (e.g. Cluster-in-a-box. Cluster-across-boxes), how the quorum and the other shared disks have to be presented to the MSCS nodes. For a detailed description take a look at VMware's Cluster configuration guide at http://kb.vmware.com/kb/1004617

Before data is written to a block on a virtual disk for the first time, the block is zeroed out. With lazy zeroed disks this means the first write to a block results in 2 write operations. To avoid this overhead, you can create the virtual disk as Eager Zeroed. In fact, this is a requirement for the MSCS configuration (see documentation).

André

[mahesh414]

I have created datastore of 50GB (LUN) and trying to create a Eagar zeroed vmdk disk of size 49GB on the same datastore.  It takes long time (approx 1 1/2 hrs) to create 5GB vmdk file and errors after completing 100%.

Is there any known issue with Eagar zeroed disk or I am doing any mistake here?  Why does it take so long time to create one vmdk file?

I need to create 33 disks for different sizes using the same procedure

please help!!

[a_p_]

Usually a 50GB disk on a FC array should take only a few minutes. The issue you are facing may be caused by the low remaining disk space on the datastore. ESX needs some of the disk space to maintain its metadata and you may also consider to leave some free disk space for e.g. snapshots and/or changed block tracking. I usually create larger LUNs/datastores with multple virtual disks on them rather than one LUN per datastore.

Out of curiosity. Do you see the same behavior with creating e.g. a 40GB virtual disk on this datastore?

André

PS:The screen shot shows a host with ESX 4.0 Update 2, however you are asking about vSphere 4.1!?

[mahesh414]

Andre, We were told to create seperate Datastore for each LUN (for exp: 1GB datastore for 1GB LUN). we are eliminating RDMs slowly from our environment. This will boost up the I/O speed as it takes seperate path for each I/O request.  If we create one large LUN with seperate vmdk's the I/O request will go through only one path instead.

We dont take snapshots for these LUNs so we dont bother about the much free space on datastores.

Please let me know if you have any other thoughts better than this setup.

Note: when i create Eager Zeroed disk it takes long time and throws this error. but if i select lazy zeroed there is no issue at all.

Thanks for all your time

[rickardnobel]

mahesh414 wrote:

Note: when i create Eager Zeroed disk it takes long time and throws this error. but if i select lazy zeroed there is no issue at all.

You are not using "Thin provisioned LUNs" on the SAN side? Which might run out of space when doing all zeroing at creation time (as with eager zeroed thick disks.)