This post is more than 5 years old
256 Posts
0
1471
Are these still good Oracle on VMW guidelines???
Matt Kaberlein raised these questions today, which seemed like good ones to me:
Just wondering if the following are still good best practices guidelines at this point:
- Dedicate specific datastores to specific Oracle data types and/or Oracle DBs?
- Map VMDKs to separate datastores to increase isolation and performance parallelism?
- Leave roughly 20% of available capacity on each datastore for ESX overhead if applicable (vMotion, VM snaps)?
kmani
1 Message
0
March 21st, 2014 18:00
Yes all are valid, adding to Jeff. 20% is generally recommended for production workloads.
jeff_browning
256 Posts
0
March 17th, 2014 10:00
I agree on everything except the 20% free space. Don would have to say if that’s still necessary, but I rather doubt it. VMFS has become much better behaved in recent releases. I doubt you need that much free space at this point. However, Don may comment further.
In terms of the isolation stuff, yes, certainly put separate Oracle databases on separate .vmdk files which are, in turn, on separate VMFS file systems. That way, you will be able to snap / clone each database individually. Otherwise, you waste IOPS and space (making additional copies of updated blocks in the non-used database).
zhaos2
2 Intern
2 Intern
•
643 Posts
1
March 24th, 2014 00:00
For the first question, a datastore is always an abstraction of the storage tier for oracle applications and, therefore, it’s a generally accepted best practice to create dedicated datastores if Oracle customers need to define individual service level guarantees for different database file (whether log or database files).
For the second question, are you saying you have different VMDK files from different Oracle virtual servers and need to map those vmdks to separate datastores? If yes, I guess what you’re saying is to separate different oracle databases by mapping them to different datastores, or maybe it’s different files from one database instead. I think, whatever it is, different datastores consist of different LUNs from storage array finally. So theoretically I believe that the mapping means more control from DBA or Storage Admin though it’s not a substitute for oracle performance.
zhaos2
2 Intern
2 Intern
•
643 Posts
0
March 24th, 2014 20:00
For the third question, if you are migrating virtual machines between two ESX hosts that store the virtual machine swap files in different locations, vCenter may require enough disk space to store two full sized copies of the virtual machine swap file. If there is not enough space in the datastore, you will receive error message.
There are no Best practices for VM snap, it depends on your VM’s change rate.