Hi Jordan, No we do not stripe or drbd/raid our OSTs but our MDT/MGS data is drbd:ed, backing storage is raid10 over 6 disks and then primary/master over two mdt/mgs machines with drbd and heartbeat. And we use it in production, if that is any help to you.
2009/3/8 Jordan Mendler <[email protected]>: > Hi all, > I did some searching and found mixed things about DRBD with Lustre. In > particular, has anyone successfully been using this configuration in a > production environment? > On paper, I would like to do something like mirroring of Thumper-like > systems, and then striping across several of these 'pairs'. > For redundancy at both the drive and system level, essentially a stack > of: (a) each server has a 'local' RAID6 (b) For every 'active' server, there > is a 2nd that serves in a DRBD network RAID1 mirror for failover of both the > OSS and the OST on the same system (through heartbeat) (c) Lustre striping > across several of these groups. > Any successful deployments like this? From my search, the limited info I can > find on DRBD with Lustre seems more experimental. > Thanks so much, > Jordan > _______________________________________________ > Lustre-discuss mailing list > [email protected] > http://lists.lustre.org/mailman/listinfo/lustre-discuss > > -- Timh Bergström System Operations Manager Diino AB - www.diino.com :wq _______________________________________________ Lustre-discuss mailing list [email protected] http://lists.lustre.org/mailman/listinfo/lustre-discuss
