we're currently evaluating using thin provisionig in a full server (no clients/VDI) environment
are there best practices and caveats?
i mean:
does thin provisioning induce fragmentation and performance degradation on vmfs datastore?
what happens when a datastore is overcommitted and a VM tries to allocate more storage? does the VM halt abruptly?
Hello.
are there best practices and caveats?
The biggest best practice and caveat both is to have additional monitoring in place to ensure that you do not run out of space. vCenter has this out of the box, but make sure it is set up to your liking.
Other caveats:
Running defrags will cause thin disks(s) to grow to varying degrees.
Can't be used for FT enabled VMs in vSphere.
does thin provisioning induce fragmentation and performance degradation on vmfs datastore?
Fragmentation is not a problem. For write intensive workloads, there is a performance impact. Read more about the performance in the Performance Study of VMware vStorage Thin Provisioning.
what happens when a datastore is overcommitted and a VM tries to allocate more storage? does the VM halt abruptly?
Read more about that in "[Easy recovery from a full VMware ESX datastore|http://www.vcritical.com/2009/10/easy-recovery-from-a-full-vmware-esx-datastore/]" at Eric Gray's site.
Good Luck!
Hello.
are there best practices and caveats?
The biggest best practice and caveat both is to have additional monitoring in place to ensure that you do not run out of space. vCenter has this out of the box, but make sure it is set up to your liking.
Other caveats:
Running defrags will cause thin disks(s) to grow to varying degrees.
Can't be used for FT enabled VMs in vSphere.
does thin provisioning induce fragmentation and performance degradation on vmfs datastore?
Fragmentation is not a problem. For write intensive workloads, there is a performance impact. Read more about the performance in the Performance Study of VMware vStorage Thin Provisioning.
what happens when a datastore is overcommitted and a VM tries to allocate more storage? does the VM halt abruptly?
Read more about that in "[Easy recovery from a full VMware ESX datastore|http://www.vcritical.com/2009/10/easy-recovery-from-a-full-vmware-esx-datastore/]" at Eric Gray's site.
Good Luck!
very interesting info
how one can hunt down erroneously thin-created disks?
is there a faster way than opening each vm and looking at disks definitions?
Yes, use the free RVTools. The vDisk tab will show you all of your thin disks.
I use almost exclusively thin provisioning, only some critical database servers have thick disks but it is not because of performance, it is beause those VMs are not allowed to stop if data store gets full. Thin disks tend to grow during their life time even if you delete files inside VM, free space is not actually reclaimed. To reclaim free space and shrink thin disks I use sdelete(*) tool with option "-c" which will wipe out free blocks on NTFS,
(*) http://technet.microsoft.com/en-us/sysinternals/bb897443.aspx
For the VM to be paused when a full datastore event occurs, I'm assuming the VMtools must be installed and up to date on the VM?
As without them, vcenter cannot pause the VM correct?
Does the pausing of the VM happen the same to all types of guest OS's or just Windows?