Clarifications needed

[babblo]

Hi!

After reading about freeNAS and ZFS on the docs and forum, I still have several noob questions.

I want to make a backup server mirrored with the ability to survive 2 simultaneous HDs failures. (Bad sectos, complete dead HDs, etc.)

If I understand, I will need 3 HDs (RAIDZ-2) for this. Correct?

Now, suppose I need to add more space, can I attach another 3 HDs and grow up the volumes online?

Sending snapshots at regular intervals to a remote machine (the idea is have a "clone" of the backup machine), in case of a complete disaster (maybe a meteorite strike my house) and loosing the complete machine, will I be able to retrieve the exact state at last snapshot? (I mean, just with the snapshot data) Or even move the "clone" backup machine to the original one and keep working as usual?


Thanks in advance, and sorry for my english, isn't my native language.


PS: In return, If my evil plans work ok, I will post all steps as "a noob guide for complete paranoid backup solution". :)

 

[ben]

4 HDDs is the minimum for RAID-Z2. If you only have three drives, you can put them in a three-way mirror. Recommended sizes for RAID-Z2 are 4, 6, or 10 drives.

ZFS RAID pools can't be grown except by replacing one disk at a time with larger ones and resilvering. You can, however, stripe in additional RAID pools later. Just realize that any group of disks striped into a pool can never be removed without total data loss. Each new group would have its own redundancy, rather than sharing in the existing one (so a second 4-drive RAID-Z2 loses another two disks to redundancy). Remote snapshot replication is built right into the GUI. If you have a viable snapshot, it can be cloned and used locally or on any ZFS pool you care to send it to.

That backup plan isn't paranoid... it's baseline best practice. ;)

 

[babblo]

Ben, thank you very much for your comments and clarifications.

As noob, I'm still discovering a lot of things about ZFS (looks absolutely awesome). After reading your comments, and a lot more from ZFS documentation, I will not go with RAID-Zn but with a mirrored schema, specifically with 3 HDs so I can resist 2 disk outages. Also, a remote server (in a remote location, obviously) with another 3 HDs receiving snapshots from the first server. This way, i should loose 6 HDs before a complete data loss. (If that happend, I promess I will change my profession to something like a farmer..)

Btw, I already did some tests using VM's and everything run as expected, but give me some more questions I was unable to find answer.

- Can I add a 4th disk to the existing 3 disk mirrored array?
- If the meteorite strike my house, having the remote snapshot, the only thing I should do is rebuild the machine and reverse the replication direction one time. I am right?
- Is there any way, from the freeNAS GUI, to see the status (progress is a plus) of, for example, a disk synchronization after a replace?

Once again, thank you.

 

[ben]

Now that's a properly paranoid plan, disk-wise at least.

I instinctively want to say that you can mirror in more disks, but I don't know how to do so, so I won't give it a definite yes. I'm also not sure what kind of feedback the Web UI gives on a ZFS scrub or a resilver, sorry.

As for rebuilding after total loss of a primary machine, all you need to do is a remote replication from the backup device to the new main one (the reverse of the one you scheduled to get the backup) using the most recent complete snapshot.

It's also good to have a way to connect to the backup device regularly, clone a snapshot there and use it like it's the original - just to check the integrity of the backup without waiting for disaster to strike.

 

[babblo]

Now that's a properly paranoid plan, disk-wise at least.

Thanks! (Saving reply... backup reply.. :P )

It's also good to have a way to connect to the backup device regularly, clone a snapshot there and use it like it's the original - just to check the integrity of the backup without waiting for disaster to strike.

Yep, I have something like this working automagically.. ("now porting" to ZFS/Snapshots in freeNAS) The idea is simple.. the backup job generate a file with a list (n% of files, for speed) of random files and their checksums[*] (location/filename;checksum), then on scheduled basis, another script check if the files listed on the previously generated file still exist and have the right checksums. ©


[*] Currently I'm using md5 but with large files (they are randomly selected) the script add a lot of time to complete, any fast alternative?

 

[paleoN]

Now that's a properly paranoid plan, disk-wise at least.

A bit foolish as well considering there is a 2[SUP]nd[/SUP] separate machine with a complete backup of the first. The OP would get want he wants with a simple mirror on both the primary & backup machines. The primary would need to be rebuilt sooner in some scenarios with a 2-way mirror.

I instinctively want to say that you can mirror in more disks, but I don't know how to do so, so I won't give it a definite yes.

You can. It just doesn't help much. From the ZFS Best Practices Guide:

If you need better data protection, a 3-way mirror has a significantly greater MTTDL than a 2-way mirror. Going to a 4-way (or greater) mirror may offer only marginal improvements in data protection. Concentrate on other methods of data protection if a 3-way mirror is insufficient.