[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [linux-lvm] xfs_freeze & lvm2



i'm sorry for breaking the thread, but i am reading responses through web archive ;)

..

>> as i understand it, consistent copy is created when lvcreate is
>> issued, not when snapshot volume is mounted - is this right ?

> As i understood things, no. Mounttime matters ;-) see
> <http://www.tldp.org/HOWTO/LVM-HOWTO/snapshotintro.html> and as example <http://www.tldp.org/HOWTO/LVM-HOWTO/snapshots_backup.html>.

well, supposedly there is no way to see wether changes are propogated to snapshot between lvcreate and mount because you must mount it to see (reminds about cats & boxes) ;)

but there are two things that lead me to believe it is lvcreate that matters :

1. "You should remove snapshot volume when you have finished with them because they take a copy of all data written to the original volume and this can hurt performance."

note, it says "remove", not unmount. additionally, you can easily mount and unmount snapshot several times - it would be pretty hard to have a consistent copy in such a case;

2. i did some testing before : created a snapshot (but did not mount it), made some modifications to original data and checked wether used space would increase for snapshot volume (with lvdisplay). well, it did - so i assumed snapshot is created, well, when it is created :)

..

>> if it actually is done together with snapshot, does this mean that i
>> really can forget about xfs_freeze and just create snapshots ?

well, what about this part ?
from what i have read here and also in other places, xfs filesystem should be consistent, but i would like to be sure (though i might try to test this by simulating activity during lvcreate, i would prefer not to)

>> must i specify ro in mount options or is this optional ?
> Well, if i'm wrong ( and to be honest, i didn't play around with all
> the stuff for 1 1/2 years now :-( ) somebody might correct me, but if
> i remember postings from linux-xfs mailing list, rw-snapshots with XFS
> is not quite unstable ...

well, as i don't need r/w anyway, i'll stick with ro, just to be sure then :)

> Cheers
> Klaus
--
 Rich


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]