vSphere 5.5 Large Datastore

[lvanek]

i have a customer who needs 10TB of space allocated to 1VM.

this is not a request i have had before so i am not sure what is the best direction to lead them as what is the best way to implement it.

Currently the largest LUN they carve on the SAN is 2TB,

Should they just create a 10TB LUN and 10TB datastore and call it a day?

or 5 x 2TB LUNs combined into 1 Datastore

or 5 x 2Tb LUNs w/ 5 x 2TB datastores configured in a storage group and the vmdk spread across those...

any input on this would be wonderful as i am sure others have had to do this already, and i just have not in 9 years of doing VMWare.

[rcporto]

The 10TB will be used by a single partition inside de guest OS ? Or it will be used by multiple different partitions ?

Anyway I recommend you avoid multi extent datastore and spread a single guest OS partition across multiple vmdk. I already see a lot of environment using this configuration but this was years ago due to the 2TB disks limitations on older vSphere versions... and depending of the size, the management effort will grow as your infrastructure grow.

So, if the 10TB will be used by a single vmdk, create 1 to 1 LUN/datastore... if not, create multiple LUN/datastores.

[lvanek]

it is one single disk to be presented to the server, i believe oracle server.

the server itself needs a nice fat 10tb drive

[rcporto]

Since you will use Oracle, my recommendation is ask the DBA if the disks will be managed by ASM (Automatic Storage Management), if yes you can create multiple LUN/datastore and create a single vmdk per datastore, present to Oracle ASM and ask the DBA to create a disk group with the multiple disks. This way, depending of your storage architecture, you can balance these LUNs by storage array processors and paths.

[lvanek]

‌jjust double checked. The oracle was a diff thing they needed.

THis this is for a riverbed wan optimizer OVF. Linux distro and does NOT support multiple disks

[rcporto]

So, create a single LUN and format as a single datastore. If you plan take snapshot, add some additional space on datastore.