ZFS Restore a part of a snapshot when low on disk space

fluca1978

fluca1978

I've a dataset with an attached snapshot.
I deleted contents from the dataset, clearly not reducing disk space usage due to the snapshot. Now I would like to move some things out of the snapshot to the original dataset (not the whole snapshot) and throw away the snapshot, but I don't have enough disk space to have the same data on the dataset and the snapshot.
Therefore, copying from
Code: 
.zfs/snapshot
is not suitable. How can I proceed?
 
T-Aoki said:
Isn't there some prerequisites (if I recall correctly, chosen checksum algo) and extreme memory consumptions?
Possibly fast_dedupe help for the latter?
Click to expand...

Well, OP didn't tell us anything about your computer. How much RAM do you have, OP?

Generally speaking dedup RAM requirements are not nearly as bad as they are said to be.

Just be sure to configure some swapspace so that you don't have too many userland kills.
 
cracauer@ said:
Well, OP didn't tell us anything about your computer. How much RAM do you have, OP?

Generally speaking dedup RAM requirements are not nearly as bad as they are said to be.

Just be sure to configure some swapspace so that you don't have too many userland kills.
Click to expand...

This an experimental virtual machine with 1GB of ram, and eight 1GB disks arranged in raidz3 pool.

I'm interested in practicing on how to solve this kind of issue, that clearly could involve moving to another storage the data I want to keep, remove the dataset and crate it again. But there could be a more zfs-ish way of doing this.
 
fluca1978 said:
I've a dataset with an attached snapshot.
I deleted contents from the dataset, clearly not reducing disk space usage due to the snapshot. Now I would like to move some things out of the snapshot to the original dataset (not the whole snapshot) and throw away the snapshot,
Click to expand...
Unless you have some other test case in mind, I would zfs-rollback(8) the dataset from the snapshot, destroy the snapshot, deleting files from the dataset, freeing space.
 
If you have older and can-be-deleted snapshots, destroying them would free some spaces for you, if there are some actual increases in space.
 
Did you happen look at the article I linked to? It seems like setting that sysctl might do the trick for you, allowing you to cheaply copy the files out of the snapshot and back to the live filesystem and then delete the snapshot. (I have not tested this explicitly, but it is easy to try and worth a shot.)

Here it is again:
Eric A. Borisch said:
I wonder if block replication will come to the rescue: https://klarasystems.com/articles/accelerating-zfs-with-copy-offloading-brt/
Click to expand...
 
Eric A. Borisch said:
Did you happen look at the article I linked to? It seems like setting that sysctl might do the trick for you, allowing you to cheaply copy the files out of the snapshot and back to the live filesystem and then delete the snapshot. (I have not tested this explicitly, but it is easy to try and worth a shot.)

Here it is again:
Click to expand...

This was literally magic, a feature I was not aware of. Thanks!
 
fjdlr said:
It's not possible to attach an external USB 3.0 hard drive.
Create a zpool and switch zdata to it?
Click to expand...

Of course it is, or better, being a virtual machine I can attach another piece of storage or grow a volume, but I was trying to fix a mess to learn how to solve in the case I cannot expand anymore.
 
fluca1978 said:
Of course it is, or better, being a virtual machine I can attach another piece of storage or grow a volume, but I was trying to fix a mess to learn how to solve in the case I cannot expand anymore.
Click to expand...
Okay, I didn't understand the situation well, sometimes I have a bad understanding when reading in English
 
You must log in or register to reply here.
Back
Top