Snapshot as Backup-Option for unwanted deletions

B

bsus

Hi,
I want to ask if it is safe to use ZFS-Snapshots to provide accidently deletions.

Following situation:
There some stupid sectarian working with important files. One of them has selected a folder and accidentally deleted some files.
Is it recommand to use the ZFS snapshot function to avoid this issue?

Regards, bsus
 
I sincerely apologise for the uselessness of this reply, but I have to ask...

bsus said:
to provide accidently deletions.
Click to expand...
Providing accidental deletions? Who are you, the BOFH? :e

Fonz (couldn't help myself, sorry)
 
Hmm, I could believe that the time comes where I press blind the wrong button.
Instead of Paste Delete, or I slid the wrong menue entry while I want to copy something.

I am sure that this isn't impossible ;)

Are there some more detailed informations how the snapshot works.
I can't beleave that this is so safe because the image is just virtualy and when I delete something it is gone even zfs doesn't overwrite this directly with zeros but by the next bigger copy-procedure should then overwrite the free given space where my deleted file could be or not?
 
bsus said:
Hi,
I want to ask if it is safe to use ZFS-Snapshots to provide accidently deletions.

Following situation:
There some stupid sectarian working with important files. One of them has selected a folder and accidentally deleted some files.
Is it recommand to use the ZFS snapshot function to avoid this issue?

Regards, bsus
Click to expand...

Actually this is the most bulletproof solution. You can also script hourly - daily snapshots so that you can prevent accidental data loss.
 
bsus said:
I can't beleave that this is so safe because the image is just virtualy and when I delete something it is gone even zfs doesn't overwrite this directly with zeros but by the next bigger copy-procedure should then overwrite the free given space where my deleted file could be or not?
Click to expand...
ZFS uses COW (copy on write) to write data to your disks. Practically this means that even if you delete a file, ZFS will not overwrite to that location. This of course can cause fragmentation in the long term.
But the bottom line is that snapshots are the safest way to protect your data.
 
Provided that you design your ZFS-filesystem to only rollback the affected data, e.g.
Code: 
NAME                          MOUNTPOINT
pool/root                     /
pool/root/home                /home
pool/root/home/secretary1     /home/secretary1
pool/root/home/secretary2     /home/secretary2
pool/root/home/secretary3     /home/secretary3


But if you rollback home and it looks like this:
Code: 
NAME                          MOUNTPOINT
pool/root                     /
pool/root/home                /home
You will rollback the homes of all secretaries. So if you have 3000 secretaries(users), you reset data for all 3000. So you will need to create zfs-filesystems for each user if you want to do it like that.

/Sebulon
 
Sebulon said:
You will rollback the homes of all secretaries. So if you have 3000 secretaries(users), you reset data for all 3000. So you will need to create zfs-filesystems for each user if you want to do it like that.
Click to expand...
It's not necessary to rollback a filesystem to recover data. Provided you know which files have changed, you can just copy the actual file(s) from the .zfs directory and overwrite the current one.
 
@carlton_draught

That definitely depends on the size. It is awesome to have users themselves restoring lost data through e.g. Previous Versions, since most restores are only a couple of MB´s, perhaps even half of that. That is the amount of data you can expect a user to restore, but if you, as an administrator is going to restore GB´s or even TB´s with copy/replace over a network, I promise you, you will wish you could do a rollback instead.

/Sebulon
 
Hmm ok, I think my actuak solution is not so good.
Actually I got a RAID6 with 6TB (5x 2TB). There two main Users:
1. One with much office files /home-dir keps about 4GB
2. Me much Film-Data, Images, very Big files not much small files - about 1,4TB

2. thinks he could renounce on snapshots.

Also the RAID6 exists out of the bad WD Caviar Green-Series

Is there a much better way?

Thanks for the help, Regards
 
Sebulon said:
@carlton_draught

That definitely depends on the size. It is awesome to have users themselves restoring lost data through e.g. Previous Versions, since most restores are only a couple of MB´s, perhaps even half of that. That is the amount of data you can expect a user to restore, but if you, as an administrator is going to restore GB´s or even TB´s with copy/replace over a network, I promise you, you will wish you could do a rollback instead.

/Sebulon
Click to expand...
I'm sure the rollback is quicker. The problem with rollbacks is that the filesystem is brought back to the state of the snapshot you are restoring to, and all intermediate snapshots must be destroyed. Fine if the delete/modification was the last change the user made. Not fine if the user made the change and then has proceeded to make other useful changes in the meantime before discovering the disaster. Rollback is a blunt tool.
 
@ carlton_draught

You have a good point. That is very important to keep in mind when choosing method. Rollback is quick but blunt.

/Sebulon
 
Hi,
1. What is the best way to make Snapshots of /home/<user> but not of /home/j?
2. In the man of zfs I read about the rollback command in relation to the snapshot.
Does this mean that when I want to recover a "missing" or "older" file I have to take the whole /home directory to this state???

Regards
 
You must log in or register to reply here.
Back
Top