Snapshotting and Replication questions

[Eniqmatic]

Hi all!

I've been setting up few servers to use a different locations. I have some files being synced to the box, snapshots setup running regularly and I have the previous versions working well and have setup replication to another FreeNAS. All seems to be well.

So I checked in this morning and the replication of 80GB had finished and I could go in to the remote FreeNAS and see a duplicate of all dataset from the first box, and the snapshots are listed under the snapshot tab. Great.

However I realized I have a couple questions. Say the first box dies and I want to restore a changed file that is in the included snapshot on the second box, what is the procedure for doing this? Do I need to clone the appropriate snapshot and restore that way or is there a way to get "previous versions" to work on the second box too? Because I do not have a snapshot task created for the replicated folder on the second box, I cannot add the tasks to the CIFS share.

Secondly, is there an unwritten rule as to how many snapshots you would want to keep? I know you can do it for as long a period as you want but I see some people saying "why would you keep that many snapshots" if it is over a few hundred. Is there a downside to keeping all snapshots? There isn't that much data being changed daily as it is a load of small files. Probably less than 50-100MB a day.

Thirdly, if the first box dies, what would your full procedure be to restore things as they were to the first site, from the second?

I apologies if things are a bit messy, I'm still quite knew to this and want to make sure I understand things. I have read the documentation on the relevant sections but couldn't see much info.

Any help appreciated as always.

 

[Ericloewe]

Secondly, is there an unwritten rule as to how many snapshots you would want to keep? I know you can do it for as long a period as you want but I see some people saying "why would you keep that many snapshots" if it is over a few hundred. Is there a downside to keeping all snapshots? There isn't that much data being changed daily as it is a load of small files. Probably less than 50-100MB a day.

Snapshots add a trivial amount of metadata. However, trivial*crapton=crapton, so do not overdo the snapshots, or else managing them will be painful.

Thirdly, if the first box dies, what would your full procedure be to restore things as they were to the first site, from the second?

Immediately? Bender is local, Flexo is backup:

Code:

*GUI export Flexo

#zpool import Flexo Bender
#zpool export Bender

*Upload original server's config to backup

 

[Eniqmatic]

Snapshots add a trivial amount of metadata. However, trivial*crapton=crapton, so do not overdo the snapshots, or else managing them will be painful.


Immediately? Bender is local, Flexo is backup:

Code:

*GUI export Flexo

#zpool import Flexo Bender
#zpool export Bender

*Upload original server's config to backup

Thanks for clearing up the snapshots confusion. What would your suggestion be for keeping these snapshots else where?

Sorry, could you elaborate on "Flexo" and "Bender" or point me in the right direction? Its the first I've heard of it and admittedly a little confused!!

Thanks for your time so far.

 

[Ericloewe]

Sorry, could you elaborate on "Flexo" and "Bender" or point me in the right

Those are the names of these two hypothetical pools.

What would your suggestion be for keeping th

I don't understand what your goal is.

You can have several snapshot schedules, say:

• Every 15 minutes, snapshots kept for a week
• Every 6 hours, snapshots kept for a month
• Every day, kept for six months

That way you can have the granularity to fix "Oh S#!@" moments that just happened while also being able to keep some versions for a much longer time, for less urgent "D'oh" moments.

 

[Revolution]

I have a little question to this topic too.

I'm using snapshots set every 15 min for a given dataset and i can see all the different versions in windows just fine.
But I have 96 snapshots per day, all with refer of 0. How can i keep this clean?

 

[Ericloewe]

I have a little question to this topic too.

I'm using snapshots set every 15 min for a given dataset and i can see all the different versions in windows just fine.
But I have 96 snapshots per day, all with refer of 0. How can i keep this clean?

You can't keep them very long, at the rate you're taking them. That's the only real solution.

 

[Eniqmatic]

Those are the names of these two hypothetical pools.


I don't understand what your goal is.

You can have several snapshot schedules, say:

• Every 15 minutes, snapshots kept for a week
• Every 6 hours, snapshots kept for a month
• Every day, kept for six months

That way you can have the granularity to fix "Oh S#!@" moments that just happened while also being able to keep some versions for a much longer time, for less urgent "D'oh" moments.

Thank you, that helps alot. Apart from the import/export thing I'm still a bit unclear but I think best thing is to setup to boxes and do it practically to see for myself.

Appreciate the help!