Re: enbd for linux diskless clients (with failover?)

Hi,
some time ago I've set up an installation where clients had built
RAID5 and/or mirror raid from nbd devices. This worked fine, so
I may imagine adding redundancy this way:
Have two servers export some chunks,and glue those chunks together
with softraid on the  client side. (Actually I've build some tree
structure
on >40 nodes -> softraid 5 built from every 8 hosts, exported to another client
which built another raid5 of those, and NFS-exported it to all the nodes :))

Regards
Michał

2005/10/9, dsuchod <[email protected]>:
>
> Hi!! I just set up a nbd server and booted 15 diskless linux clients from
> it. I was quite impressed from the performance (better and more efficient
> than my NFSv3 experience)!
>
> After reliability and performance solved there would be next question of
> interest to me - poor mans high availability: Would it be possible to put
> two equal servers exporting just the same partion with the same content to
> the clients, so if one of the server crashes the client is not doomed to
> die too. The task is not to sync writes (the clients will see their
> filesystems readonly), but just to get redundancy in cases of failing
> of one crucial device ...
>
> I would imagine some triangle setup:
>
>
>           master                          slave
>           server                          server
>         * raid1 of local partition ---  * exporting local
>         and nb0 (from slave)            partition (rw) to master
>                                         via enbd
>         * exporting local disk          * exporting local disk
>         to clients readonly             to clients ro
>                 \                           /
>                   \                       /
>                     \                   /
>                         client 1 ... N (N large ++100)
>                         * importing with one enbd
>                         client both servers (ro)
>                         * (enable local caching)
>
> Would that be possible (and stable)?
>
> By now I was mounting ramdisk to directories which should be writeable.
> Other option would be unionfs (which is not production stable) or cowloop
> (not tested yet in nbd setup). I read that enbd is able to do local caching
> - where does that cache would go (into a tempfs in RAM ...)?
>
> Thank you for any input!!
>
>         Dirk
> _______________________________________________
> ENBD mailing list
> [email protected]
> http://lists.community.tummy.com/mailman/listinfo/enbd
>

_______________________________________________
ENBD mailing list
[email protected]
http://lists.community.tummy.com/mailman/listinfo/enbd