Hello, Is it possible to setup two or more Lustre clusters -- each with their own OST''s and MDS'' -- and have clients mount these two Lustre clusters in a RAID-1 configuration? Has anyone tried mirroring Lustre at the cluster level in a way that is transparent to client applications? I''m evaluating different methods of increasing Lustre availability and reliability without using shared storage. I have some DRBD OST pairs setup, but recovering from a split-brain requires manual intervention from the administrator. So, this does not seem like it would scale well in a large cluster. Configuring two or more Lustre clusters in a RAID-1 setup seems like an interesting option. Thanks in advance, Eser
Hi This is not possible at the moment. In the 2008/2009 timeframe we hope to tackle issues like this for metadata. In 2007 mirroring OST data should become possible. - Peter -> -----Original Message----- > From: lustre-discuss-bounces@clusterfs.com > [mailto:lustre-discuss-bounces@clusterfs.com] On Behalf Of EKC > Sent: Sunday, June 25, 2006 7:49 AM > To: lustre-discuss@clusterfs.com > Subject: [Lustre-discuss] RAID1 Over Two Lustre Filesystems? > > Hello, > > Is it possible to setup two or more Lustre clusters -- each > with their own OST''s and MDS'' -- and have clients mount these > two Lustre clusters in a RAID-1 configuration? Has anyone > tried mirroring Lustre at the cluster level in a way that is > transparent to client applications? > > I''m evaluating different methods of increasing Lustre > availability and reliability without using shared storage. I > have some DRBD OST pairs setup, but recovering from a > split-brain requires manual intervention from the > administrator. So, this does not seem like it would scale > well in a large cluster. Configuring two or more Lustre clusters in a > RAID-1 setup seems like an interesting option. > > Thanks in advance, > Eser > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss@clusterfs.com > https://mail.clusterfs.com/mailman/listinfo/lustre-discuss > >
Hi, And were it totally crazy to run Lustre over nbd? I.e lets'' assume two MDS/OST machines both with disk hd0 shared as nbd0 and on machine0: raid1(local-hd0, machine1-ndb0). On machine1 the raid1 were defined but not enabled. If/when machine0 fails machine1 would enable raid1 and start the failed service. Or should we better wait for 2007? ;-) Best regards, Jozsef -- E-mail : kadlec@sunserv.kfki.hu, kadlec@blackhole.kfki.hu PGP key: http://www.kfki.hu/~kadlec/pgp_public_key.txt Address: KFKI Research Institute for Particle and Nuclear Physics H-1525 Budapest 114, POB. 49, Hungary
No, that''s not so crazy, several people run Lustre over DRBD. - Peter -> -----Original Message----- > From: lustre-discuss-bounces@clusterfs.com > [mailto:lustre-discuss-bounces@clusterfs.com] On Behalf Of > Kadlecsik Jozsi > Sent: Monday, June 26, 2006 3:10 AM > To: lustre-discuss@clusterfs.com > Subject: Re: [Lustre-discuss] RAID1 Over Two Lustre Filesystems? > > Hi, > > And were it totally crazy to run Lustre over nbd? > > I.e lets'' assume two MDS/OST machines both with disk hd0 > shared as nbd0 and on machine0: raid1(local-hd0, > machine1-ndb0). On machine1 the raid1 were defined but not > enabled. If/when machine0 fails machine1 would enable > raid1 and start the failed service. > > Or should we better wait for 2007? ;-) > > Best regards, > Jozsef > -- > E-mail : kadlec@sunserv.kfki.hu, kadlec@blackhole.kfki.hu PGP > key: http://www.kfki.hu/~kadlec/pgp_public_key.txt > Address: KFKI Research Institute for Particle and Nuclear Physics > H-1525 Budapest 114, POB. 49, Hungary > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss@clusterfs.com > https://mail.clusterfs.com/mailman/listinfo/lustre-discuss > >