iscsi and geom mirror - stupid idea or not ?

Oliver Brandmueller ob at e-Gitt.NET
Thu Apr 19 11:04:52 UTC 2007


Hi,

On Thu, Apr 19, 2007 at 10:48:16AM +0100, Pete French wrote:
> what would happen if I made a machine which contained a mirrored
> geom pair consiting of one local driive and one drive accessed via iscsi on
> a remote machine ? would this work ?
> 
> what I am considering is two such machines, geographicly distinct. one is
> a 'master' and boots off the mirrored drive, the other is a slave and
> has a separate boot drive which just rngs FreeBSD to make the drive inside
> it into an iscsi target for the first machine. The idea here is if the first
> machine is catastrphicly killed (like building falls down on it or
> something) then the second one can be rebooted from the internal drive, and
> will hence become the first one. It's basically a way of making a standby
> machine in case of disaster.
> 
> I havent really looked at iSCSI until recently, and this is just one of
> the ideas I came up with looking at the possibilities.

You could also go and use ggate for that. And seems to get more and more 
common to work like that, although probably most setups I heard of 
probably don't have a long distance link between them.

There are a few things you should consider: First, you have to make 
absolutely sure, that for example the mirrored disk is not attached if 
after a crash the original master comes back and the slave took over. If 
this happens you're likely to damage something really bad.

Second is, something like this gives you mirrored data with practically 
no gap to the original disk. The price you have to pay: This does not 
help you against logical errors (a filesystem damage will be replicated 
just fine...). A setup like this does not serve as a backup.

Third is, you'll have to fsck everything, so this defines your minimum 
service outage. I'm not sure, if I'd trust background fsck here, also bg 
fsck is a big performance penalty, which might or might not be a 
problem for your setup.


A replication (like rsync, ssync or similar) sure has the drawback of 
the replication gap for the data. Also you cannot just take over the IP 
of the NFS server, but have to remount everything. But you have fsck 
time, lower chance damage due to logical error and the nice effect, that 
you could do your backups from the replicated data, not affecting your 
live system. But have to deal with lost data from probably several hours 
or how to replicate changed data after recovery.


- Olli


-- 
| Oliver Brandmueller | Offenbacher Str. 1  | Germany       D-14197 Berlin |
| Fon +49-172-3130856 | Fax +49-172-3145027 | WWW:   http://the.addict.de/ |
|               Ich bin das Internet. Sowahr ich Gott helfe.               |
| Eine gewerbliche Nutzung aller enthaltenen Adressen ist nicht gestattet! |
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20070419/710c5060/attachment.pgp


More information about the freebsd-stable mailing list