Any best practises for dataset layout for use w/o recursive replication?

[Roger Wilco]

Hello,

as there seem to be issues with recursive replication which might lead to it's removal (see jkh's comment in ticket #2360 or search e.g. for freenas:state in the bug tracker) I am wondering how to layout the datasets in the zpool properly.

The whole system is still a lab installation, so I am not bound to any constrains.

Right now the layout basically looks like this:

Code:

pool
  cifs
    cifs/shares
      cifs/shares/public
      cifs/shares/scanner
      cifs/shares/transfer
    cifs/users
      cifs/users/user_1
      cifs/users/user_n
  jails
    jails/.warden-template...
    jails/jail_1
    jails/jail_n
  multimedia
    multimedia/audio
    multimedia/video
  timemachine
  postgres

My plan has been to setup recursive replication for cifs and multimedia and replicate jails, timemachine and postgres datasets manually (while switched off, or through Postgres' replication).

In case the recursive replication gets removed in FreeNAS I could set up cron jobs and still replicate recursively, but I'd rather like to stick with the GUI.

This would probably mean that I would have to set up one snapshot and replication task for each of the public, scanner, transfer, user_1, ..., user_n, audio, video datasets. But in this case it would be easier to have a completely flat layout (without the cifs, cifs/shares, cifs/users, multimedia datasets). I don't like this idea very much....

So how do you layout your filesystems?

Thanks and bye,
RW

 

[cyberjock]

This question has been asked a bunch recently. The answer is always the same... It really depends on how *you* want to break it up. Datasets are considered file system boundaries, so they bring advantages (and disadvantages) to the table by using them liberally. As long as *you* are happy with the organization (and the limitations of said organization) that's all that really matters.

 

[Roger Wilco]

Yes, you are right.
That's why I didn't ask how I should layout my datasets, but how others have done it (best practices).

-- Roger

 

[SweetAndLow]

Code:

Pool
    media(dataset)
        movies(directory)
        tv-shows(directory)
        music(directory)
    homes (dataset)
        user_1(directory)
        user_2(directory)
    jails (dataset)
        Plex
        Crashplan

I have things broken up in order of importance and how often data changes in that area. I use the homes feature of smb so my user home directory shares are taken care of. Then I have a single share for the media dataset. My desired needs are super simple and if it's complicated the user can just use their home directory.