Re: Nested File System Dedup Question
- From: E Bridenbecker <
- Subject: Re: Nested File System Dedup Question
- Date: Sun, 21 Jul 2013 06:43:22 -0500
Okay, so then when I'm calculating the amount of RAM I need, do I need
to calculate based on using dedup for the entire zpool?
On Sun, Jul 21, 2013 at 6:37 AM, Andrew Gabriel
> E Bridenbecker wrote:
>> Please bear with me as I try to put my question into words. At least
>> words that someone else can understand.
>> I'm running Solaris 11.1. I have my primary storage pool. Inside that
>> I made a file system for iSCSI targets and turned dedup on. At this
>> point I'm not sure what best practice is for creating thin provisioned
>> iSCSI targets. I have read several ways of doing it, and it seems like
>> the best way is to create another nested ZFS child under that.
>> So here's what I have done so far:
>> zpool create storagepool raidz2 ....
>> zfs create storagepool/iscsitargets
>> zfs set dedup=on storagepool/iscsitargets
>> zfs create storagepool/iscsitargets/linuxvm01
>> zfs create storagepool/iscsitargets/linuxvm02
>> zfs create storagepool/iscsitargets/windowsvm01
>> zfs create storagepool/iscsitargets/windowsvm02
>> My question is:
>> windowsvm01 & windowsvm02 have probably close to 98% exact same data
>> on them. Am I going to see the benefit of the parent
>> (storagepool/iscsitargets) file system having dedup on or is it only
>> going to autonomously dedup within windowsvm01 & windowsvm02
> Although enabled on a per-dataset basis (a filesystem or a zvol),
> deduplication of blocks happens across the whole zpool, so yes, identical
> blocks across windowsvm01 and windowsvm02 will be deduped.
> Andrew Gabriel