Optimal ZFS Volume/Dataset

Status
Not open for further replies.

ESDC

Dabbler
Joined
Oct 16, 2013
Messages
16
Hi Everyone,

Just wanted to ask a quick question what the optimal configuration is for ZFS Volumes and Datasets. How did you setup yours?

1) We have 6 x 1TB in RaidZ2
2) Here is how we arranged the Dataset:
ZFS_volume
> dataset_AFP​
>> AFP_MBP1​
>> AFP_MBP2​
> dataset_CIFS​
>> CIFS_All_Workstations​
>> CIFS_Software​
>> CIFS_ISO​
> dataset_NFS​
>> NFS_Workstations​
>> NFS_Public​
>> NFS_Server1​
>> NFS_Server2​
>> NFS_Server3​
Comments?
 

Dusan

Guru
Joined
Jan 29, 2013
Messages
1,165
There is no need to create datasets for every network share. You can just create directories and share those.
The are few reasons to use datasets:
  • You can only snapshot & replicate entire volumes or datasets
  • You can set different ZFS properties for datasets (quota, compression, reservation, ...)
However datasets also have drawbacks. For example as datasets are separate filesystems, to move a file from a dataset to dataset it needs to be copied (slow), while moving files between directories is just a simple metadata update (fast). So, I would not create datasets unless my snapshot strategy needs them or I need different ZFS properties.
 

indy

Patron
Joined
Dec 28, 2013
Messages
287
That is good to know Dusan.
When I use CIFS to move a file inside a dataset but between two shares, does it copy or move the file?
 

Dusan

Guru
Joined
Jan 29, 2013
Messages
1,165
It will be copied over the network -- the CIFS client will download the file form the source share and upload to the destination share.
 

ESDC

Dabbler
Joined
Oct 16, 2013
Messages
16
There is no need to create datasets for every network share. You can just create directories and share those.
The are few reasons to use datasets:
  • You can only snapshot & replicate entire volumes or datasets
  • You can set different ZFS properties for datasets (quota, compression, reservation, ...)
However datasets also have drawbacks. For example as datasets are separate filesystems, to move a file from a dataset to dataset it needs to be copied (slow), while moving files between directories is just a simple metadata update (fast). So, I would not create datasets unless my snapshot strategy needs them or I need different ZFS properties.

Thanks for the reply Dusan. Will keep that in mind.
 
Status
Not open for further replies.
Top