[NCLUG] zfs (my hard drive crashed, time to upgrade the fs?)

John Gilmore j.arthur.gilmore at gmail.com
Sat Jan 4 16:11:34 MST 2014


Unfortunately, zfs-auto-snapshot isn't available right now:

Reading package lists... Done
Building dependency tree
Reading state information... Done
Package zfs-auto-snapshot is not available, but is referred to by
another package.
This may mean that the package is missing, has been obsoleted, or
is only available from another source


Presumably, it will become available again later, and is just
temporarily gone. I didn't know it was a debian package though, that
certainly makes things simpler.

I was aware of the "zfs set copies=" thing, but does it assure that
the copies will be spread amoung the available spindles? I thought you
needed to configure raidz or some such to assure that.

Another thought occurs to me though. Would it be useful to abuse zfs
by giving it a networked block device? My brother lives next door, and
we've got a fast link, I could concievably setup a server there, and
BAM! instant offsite (sorta) snapshots!

Of course, fast is relative, and it'd be WAY slower than any other
device on the system. What kind of performance hit would that be?

On Sat, Jan 4, 2014 at 3:37 PM, Zak Smith <zak at computer.org> wrote:
> On Sat, Jan 04, 2014 at 03:27:30PM -0700, John Gilmore wrote:
>> I'd like automatic snapshots instead of the rsync dance I was doing
>> before. How do I set that up? I'll still need rsync for the remote
>> comptuers certainly, but I think that zfs is a much more reliable and
>> finer-grained solution for local (server) snapshots
>
> I don't know anything about running ZFS on LUKS or the sparse file
> tricks.  If it were my I'd create a fresh boot/root installation of
> Debian 7.3 (or whatever preferred distribution) and then copy my files
> from the old drive somewhere.   Then I'd create the zpool from scratch
> and do it the "clean" way with the old 2TB and the new one.
>
> Also note that you don't "have to" use a mirror or raidz with ZFS.
> You can simply use a single volume and get its filesystem benefits.
> Then you'd want to use a second volume as a separate backup.  That
> said, the zfs mirrors and raidz's are very reliable in terms of
> weathering various failures. The only downside is that there is
> additional "rescue" overhead if things really go bad.
>
> As for the snapshots, just use zfs-auto-snapshot.  It's a debian
> package.  It consists of a script ("zfs-auto-snapshot") and a bunch of
> files in /etc/cron* to schedule the various time-based snapshots.
>
> You'll also want to schedule a "zfs scrub" about once a week.  This
> will detect (and correct if possible) silent data corruption on pools.
>
>> I'd like precious files (/var/local/pics, /home/*/docs, a few others)
>> to have two copies on each spindle. How can I assure that?
>
> Check the man pages on "zfs set copies="
>
>
>
>
> --
> # Zak Smith    mobile 970-232-4468
>
> _______________________________________________
> NCLUG mailing list       NCLUG at lists.nclug.org
>
> To unsubscribe, subscribe, or modify
> your settings, go to:
> http://lists.nclug.org/mailman/listinfo/nclug


More information about the NCLUG mailing list