backups

dean.viens

Dabbler
Joined
Feb 23, 2023
Messages
32
Hi all,
i know mirrors are not backups so I'm trying to create a solution for backing up my whole server. i have 3 mirrors 6tb, and one mirror 4tb. i was thinking of just creating one pool single hard drive just for snapshots of each pool, and creating an offsite raid to transfer said backups to it.
is this possible? if so what size of hard drive do I need to temporarily store the snapshots for transfer via internet to offsite raid for the snapshot backups
 

danb35

Hall of Famer
Joined
Aug 16, 2011
Messages
15,504
creating one pool single hard drive just for snapshots of each pool
Snapshots don't work that way--they're stored on the pool (and dataset) to which they belong.
 

dean.viens

Dabbler
Joined
Feb 23, 2023
Messages
32
Snapshots don't work that way--they're stored on the pool (and dataset) to which they belong.
ok so then how does one create backups of their data?
 

danb35

Hall of Famer
Joined
Aug 16, 2011
Messages
15,504
ok so then how does one create backups of their data?
Lots of ways--cloud sync tasks, rsync to another system, and ZFS replication are among them. In the latter, snapshots are taken and then replicated to another pool, which may or may not be on the same system. So you could create a single-disk pool, replicate snapshots to that, and also replicate them offsite, but the snapshots themselves live on the source pool.
 

dean.viens

Dabbler
Joined
Feb 23, 2023
Messages
32
Lots of ways--cloud sync tasks, rsync to another system, and ZFS replication are among them. In the latter, snapshots are taken and then replicated to another pool, which may or may not be on the same system. So you could create a single-disk pool, replicate snapshots to that, and also replicate them offsite, but the snapshots themselves live on the source pool.
so the snapshot way means it wont be the snapshots themselves that are backed up but the data itself as a whole so the whole data amount ex 5TB backed up instead of the snapshots?
 

danb35

Hall of Famer
Joined
Aug 16, 2011
Messages
15,504
There's no way for the snapshots to exist independently from the data they represent. Thus, "sending a snapshot" results in the data referenced by that snapshot being present on the target.
 

artlessknave

Wizard
Joined
Oct 29, 2016
Messages
1,506
there is no way to make a backup that doesn't have a copy of the data. your backup destination must be large enough for the source (or a subset if you selectively backup).
replicating snapshots to another system is a relatively simpler way to make backups.
you can also send a copy to a cloud location.
 

dean.viens

Dabbler
Joined
Feb 23, 2023
Messages
32
there is no way to make a backup that doesn't have a copy of the data. your backup destination must be large enough for the source (or a subset if you selectively backup).
replicating snapshots to another system is a relatively simpler way to make backups.
you can also send a copy to a cloud location.
so it means regardless of using snapshot method, the backup destination needs to be big enough to store all the source data right?
 

artlessknave

Wizard
Joined
Oct 29, 2016
Messages
1,506
so it means regardless of using snapshot method, the backup destination needs to be big enough to store all the source data right?
I. um. I thought that's what I just said. but yes. if you want to backup 6TB, you need 6TB to put it on.
the backup doesn't have to be the most resilitant thing, since it's creating a redunant copy of the data already, so if you are ok with a bit more risk for less cost, you can do things with a bit less care (like consumer mobos - even a corrupted copy is better than no copy in the relatively unlikely chance of the mail pool actually dying), and you don't need performance at all (mine runs raidz2 wider than normally recommended)

snapshots are always worth having. they let the pool revert versions. a decent snapshot layout can make things like ransomware a joke. you basically press a button and the ransomware becomes irrelevant.
 

dean.viens

Dabbler
Joined
Feb 23, 2023
Messages
32
I. um. I thought that's what I just said. but yes. if you want to backup 6TB, you need 6TB to put it on.
the backup doesn't have to be the most resilitant thing, since it's creating a redunant copy of the data already, so if you are ok with a bit more risk for less cost, you can do things with a bit less care (like consumer mobos - even a corrupted copy is better than no copy in the relatively unlikely chance of the mail pool actually dying), and you don't need performance at all (mine runs raidz2 wider than normally recommended)

snapshots are always worth having. they let the pool revert versions. a decent snapshot layout can make things like ransomware a joke. you basically press a button and the ransomware becomes irrelevant.
is it possible to just use one large drive to backup multiple pools then or does one pool need to be backed up to a separate drive for that pool only then?
 

danb35

Hall of Famer
Joined
Aug 16, 2011
Messages
15,504

artlessknave

Wizard
Joined
Oct 29, 2016
Messages
1,506
is it possible to just use one large drive to backup multiple pools
absolutely. it all depends on how you configure it.
some people really like the services like backblaze. they can be really convenient for backups, though restores can be slow.

rsync.net also supports zfs replication. can be expensive but for some people can be better than hosting a second server.
 

Arwen

MVP
Joined
May 17, 2014
Messages
3,611
Here are some hints of how I make local single disk backups:
 

acp

Explorer
Joined
Mar 25, 2013
Messages
71
ok so then how does one create backups of their data?

For me, I have a LTO5 tape library. So I configured a bacula system. For TrueNAS Core, I create a jail and map the datasets into the jail. So far I only done backups from Core. I have done restores to a TrueNAS Scale. In Scale I manually added the package (Yes I know, that you are not suppose to do that. I only configured bacula-fd. The director and sd are in a VM running in XCP-NG with a iSCSI to SAS bridge). Once I completed my move to Scale, then I will test backing up from bacula using the frowned upon method (maybe bacula-fd can be included in scale as an optional service, wink wink).
 
Top