Hi everyone :) I would like to ask you a simple question, i''m not able to find the answer on the lustre documentation. I know its possible to strip a file across multiple OSTs and make some RAID 0. But is it possible to set up some kind of RAID 5 over OSTs. I mean one disk on each OSTs, the file is stripped over them and still fault tolerant if a node goes down. Thanks a lot for your help ;) Have a good day, Kieran
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 kieran JOYEUX wrote:> I know its possible to strip a file across multiple OSTs and make some > RAID 0. But is it possible to set up some kind of RAID 5 over OSTs. I > mean one disk on each OSTs, the file is stripped over them and still > fault tolerant if a node goes down.You are looking for ''Multi-server file declustering (RAID)''. Accoring to the roadmap (http://www.clusterfs.com/roadmap.html) it is planned for version 2.0 (Q3 - 2008). Lennard -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (MingW32) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFG7kcQB3IFhTJpAVkRAmmIAJ40rmx5vVSQJhnWLhoKzRlMCAYLcgCg6o7D BlES9m0HUS0rzu84fjMJWac=2aa4 -----END PGP SIGNATURE-----
Thanks a lot for your reply Lennard. I guess I''ll have a look at it on Q3 2008 ;) Goodbye Lennard Bakker wrote:> -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > kieran JOYEUX wrote: > >> I know its possible to strip a file across multiple OSTs and make some >> RAID 0. But is it possible to set up some kind of RAID 5 over OSTs. I >> mean one disk on each OSTs, the file is stripped over them and still >> fault tolerant if a node goes down. >> > > You are looking for ''Multi-server file declustering (RAID)''. > Accoring to the roadmap (http://www.clusterfs.com/roadmap.html) it is > planned for version 2.0 (Q3 - 2008). > > Lennard > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.4.6 (MingW32) > Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org > > iD8DBQFG7kcQB3IFhTJpAVkRAmmIAJ40rmx5vVSQJhnWLhoKzRlMCAYLcgCg6o7D > BlES9m0HUS0rzu84fjMJWac> =2aa4 > -----END PGP SIGNATURE----- > >
One more thing Lennard :) When i read that Lustre is a fault tolerant FS, is it thanks to Heartbeat or something else ? Sorry if my question sounds dumb... :) Thanks !! :) Lennard Bakker wrote:> -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > kieran JOYEUX wrote: > >> I know its possible to strip a file across multiple OSTs and make some >> RAID 0. But is it possible to set up some kind of RAID 5 over OSTs. I >> mean one disk on each OSTs, the file is stripped over them and still >> fault tolerant if a node goes down. >> > > You are looking for ''Multi-server file declustering (RAID)''. > Accoring to the roadmap (http://www.clusterfs.com/roadmap.html) it is > planned for version 2.0 (Q3 - 2008). > > Lennard > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.4.6 (MingW32) > Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org > > iD8DBQFG7kcQB3IFhTJpAVkRAmmIAJ40rmx5vVSQJhnWLhoKzRlMCAYLcgCg6o7D > BlES9m0HUS0rzu84fjMJWac> =2aa4 > -----END PGP SIGNATURE----- > >
On 9/17/07, kieran JOYEUX <kjoyeux at jouy.inra.fr> wrote:> One more thing Lennard :) When i read that Lustre is a fault tolerant > FS, is it thanks to Heartbeat or something else ? Sorry if my question > sounds dumb... :)Heartbeat is component to detect failures, It will discover faults related to node(s) crash and will trigger recovery process. There are lots of components which participate in recovery process e.g. journals, DLM, meta data server , raid, etc. Synchronized co-ordination between these components will recover user files. --pravin.> > Thanks !! :) > > Lennard Bakker wrote: > > -----BEGIN PGP SIGNED MESSAGE----- > > Hash: SHA1 > > > > kieran JOYEUX wrote: > > > >> I know its possible to strip a file across multiple OSTs and make some > >> RAID 0. But is it possible to set up some kind of RAID 5 over OSTs. I > >> mean one disk on each OSTs, the file is stripped over them and still > >> fault tolerant if a node goes down. > >> > > > > You are looking for ''Multi-server file declustering (RAID)''. > > Accoring to the roadmap (http://www.clusterfs.com/roadmap.html) it is > > planned for version 2.0 (Q3 - 2008). > > > > Lennard > > -----BEGIN PGP SIGNATURE----- > > Version: GnuPG v1.4.6 (MingW32) > > Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org > > > > iD8DBQFG7kcQB3IFhTJpAVkRAmmIAJ40rmx5vVSQJhnWLhoKzRlMCAYLcgCg6o7D > > BlES9m0HUS0rzu84fjMJWac> > =2aa4 > > -----END PGP SIGNATURE----- > > > > > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss at clusterfs.com > https://mail.clusterfs.com/mailman/listinfo/lustre-discuss > >
Thanks for your reply ;) So every OSTs, MGS or MDTs are SPOFs and needs to be covered from crashing. Each of them must have a passive node to fail over to... That involve a pretty complicated architecture! Pravin Shelar wrote:> On 9/17/07, kieran JOYEUX <kjoyeux at jouy.inra.fr> wrote: > >> One more thing Lennard :) When i read that Lustre is a fault tolerant >> FS, is it thanks to Heartbeat or something else ? Sorry if my question >> sounds dumb... :) >> > > Heartbeat is component to detect failures, It will discover faults > related to node(s) crash and will trigger recovery process. There are > lots of components which participate in recovery process e.g. > journals, DLM, meta data server , raid, etc. > Synchronized co-ordination between these components will recover user files. > > > --pravin. > > >> Thanks !! :) >> >> Lennard Bakker wrote: >> >>> -----BEGIN PGP SIGNED MESSAGE----- >>> Hash: SHA1 >>> >>> kieran JOYEUX wrote: >>> >>> >>>> I know its possible to strip a file across multiple OSTs and make some >>>> RAID 0. But is it possible to set up some kind of RAID 5 over OSTs. I >>>> mean one disk on each OSTs, the file is stripped over them and still >>>> fault tolerant if a node goes down. >>>> >>>> >>> You are looking for ''Multi-server file declustering (RAID)''. >>> Accoring to the roadmap (http://www.clusterfs.com/roadmap.html) it is >>> planned for version 2.0 (Q3 - 2008). >>> >>> Lennard >>> -----BEGIN PGP SIGNATURE----- >>> Version: GnuPG v1.4.6 (MingW32) >>> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org >>> >>> iD8DBQFG7kcQB3IFhTJpAVkRAmmIAJ40rmx5vVSQJhnWLhoKzRlMCAYLcgCg6o7D >>> BlES9m0HUS0rzu84fjMJWac>>> =2aa4 >>> -----END PGP SIGNATURE----- >>> >>> >>> >> _______________________________________________ >> Lustre-discuss mailing list >> Lustre-discuss at clusterfs.com >> https://mail.clusterfs.com/mailman/listinfo/lustre-discuss >> >> >> > > >