There is a lot more about snapshots and backup schemes that I don't
know than there is that I do know...  My needs are pretty simple and
small compared to some of the heavy users here, but I'd like to run
just the rough idea by the group.

One backup I need to make is from to linux online servers to a zpool.

Whats involved is two websites I use sometimes quite a bit but more
often not so much and sometimes not at all for quite a stretch.

I thought I'd rsync  the base directory to a zpool using the `--inplace'
flag to rsync that has been discussed here a time or two.

It appears the snapshot setup was including the specific zpool I plan
to use but at a frequency level way above what seems I'd need.  But them
I'm not really sure how to go at this.

I see mnthly weekly hourly and frequent which appear to be every 15
min. 

I thought it might be better to turn that off completely.  Then use a
cron job to run the rsync once per week.   The same little script
would run a snapshot following the completion of the rsync run.

I'd probably use something like this for rsync:

Cron calls  my script and runs:
  rsync -avvz --stats --delete --exclude-from='exclude_file'
            --delete-excluded --inplace 
             m...@remote:/home/me/public_html/ /www/remote_www/

It would be run with the `authorized_keys' mechanism
  
Then runs a snapshot of zfs z3/www

I'm thinking that snapshot would catch every thing since the last run.
   
Is that a reasonable plan?  

Sorry to ruminate on here... but would like any ideas you may have.

I'd probably do something very similar to the www plan with the home
lan machines... Two linux boxes are no problem, but I haven't really
figured out the two windows XP machines yet.  It seems there's always
been bugs in Cifs implementation preventing reliable
connections.. Currently I can only see top level of any windows shares
from zfs host.

A while back I had `retrospect' (a windows version) connecting to zfs
machine, but that fell through so often... it was not working.

Windows based backup tools (at least some of them) expect the admin to
be able to navigate to appropriate shares or at least use UNK
addressing.   But I have yet to get the zfs machine to show up in
windows network places.  Even when I can force things by typing an UNK
address on windows explorer.... there is a very good chance after
either machine is rebooted ... the connection would not work thereafter.

Seems like initiating the connections from zfs host is a better way to
go.  And I'm currently looking in to using Bacula.

Using rsync against the windows machines seems to mostly be a
non-starter due to something in ntfs or the OS that makes files appear
to be all new every so often causing rsync to be fooled into pulling
the same files across again.

That, by itself is bad enough but not sure what impact it might have
on zfs snapshots. .. maybe none... long as --inplace is being used.

I've been pretty lazy about it and have continued to backup the
windows machines with things like ghost... making complete drive images.

But since it seems like nothing but headaches trying to connect to the
zfs host reliably...(From windows), I run the backup onto an internal
drive and then have to move the monster file again to get it on a zfs
filesystem.

_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

Reply via email to