On Apr 12, 2005 22:42 +0400, Andrew Velikoredchanin wrote:> I want install Lustre FS in ouer Linux claster. But we need duplicate > data. I read about replication on Lustre, but I do not undestend what > need doing for this. Please, answer - what need for it?Lustre does not currently support replication - this is a feature planned for the future. Currently, systems run with some sort of RAID for the back-end storage to provide redundancy in case of disk failure. If the RAID can be attached to multiple nodes (e.g. FC, or multiport SCSI) then it is also possible to provide server failover in case of node failure. Cheers, Andreas -- Andreas Dilger Principal Software Engineer Cluster File Systems, Inc.
On Apr 13, 2005 13:52 +0400, Andrew Velikoredchanin wrote:> Andreas Dilger wrote: > >On Apr 12, 2005 22:42 +0400, Andrew Velikoredchanin wrote: > >>I want install Lustre FS in ouer Linux claster. But we need duplicate > >>data. I read about replication on Lustre, but I do not undestend what > >>need doing for this. Please, answer - what need for it? > > > >Lustre does not currently support replication - this is a feature planned > >for the future. > > > >Currently, systems run with some sort of RAID for the back-end storage to > >provide redundancy in case of disk failure. If the RAID can be attached > >to multiple nodes (e.g. FC, or multiport SCSI) then it is also possible > >to provide server failover in case of node failure. > > OK. And how RAID create in Lustre?Lustre itself will currently only do RAID0 (striping) across storage nodes (OSTs), though we have OST RAID1 support underway for a later release. Lustre uses regular block devices to hold the back-end filesystems, so it can use any kind of device that Linux supports (e.g. LVM/DM, MD RAID, FC/SCSI/SATA hardware RAID, etc). If the back-end storage is redundant then it can handle a disk failure internally. If that storage is also capable of connecting to multiple nodes (e.g. FC) then Lustre can be configured to handle node failures (though at the current time this involves some effort on the part of the implementer). Cheers, Andreas -- Andreas Dilger Principal Software Engineer Cluster File Systems, Inc.
Andreas Dilger wrote:> On Apr 12, 2005 22:42 +0400, Andrew Velikoredchanin wrote: > >>I want install Lustre FS in ouer Linux claster. But we need duplicate >>data. I read about replication on Lustre, but I do not undestend what >>need doing for this. Please, answer - what need for it? > > > Lustre does not currently support replication - this is a feature planned > for the future. > > Currently, systems run with some sort of RAID for the back-end storage to > provide redundancy in case of disk failure. If the RAID can be attached > to multiple nodes (e.g. FC, or multiport SCSI) then it is also possible > to provide server failover in case of node failure.OK. And how RAID create in Lustre?
Hi! I want install Lustre FS in ouer Linux claster. But we need duplicate data. I read about replication on Lustre, but I do not undestend what need doing for this. Please, answer - what need for it? Thanks.
On Thu, Apr 14, 2005 at 02:40:39PM -0600, Andreas Dilger wrote:> On Apr 14, 2005 11:06 -0500, Paul Jimenez wrote: > > Is it possible to use Lustre in conjunction with GNBD to obtain > > mirrored-failover capability? > > Several people have reported trying this, but I am not aware of conclusive > results either way. One issue is whether GNBD preserves the ordering > of IO requests to keep the ext3 journaling happy and the filesystem > consistent in case of a crash.I don''t know about GNBD, but DRDB preserves ordering, or preserves relevant ordering. (when configured correctly). -- Ragnar Kjørstad Software Engineer Scali - http://www.scali.com Scaling the Linux Datacenter
Andreas Dilger wrote:> On Apr 13, 2005 13:52 +0400, Andrew Velikoredchanin wrote: > >>Andreas Dilger wrote: >> >>>On Apr 12, 2005 22:42 +0400, Andrew Velikoredchanin wrote: >>> >>>>I want install Lustre FS in ouer Linux claster. But we need duplicate >>>>data. I read about replication on Lustre, but I do not undestend what >>>>need doing for this. Please, answer - what need for it? >>> >>>Lustre does not currently support replication - this is a feature planned >>>for the future. >>> >>>Currently, systems run with some sort of RAID for the back-end storage to >>>provide redundancy in case of disk failure. If the RAID can be attached >>>to multiple nodes (e.g. FC, or multiport SCSI) then it is also possible >>>to provide server failover in case of node failure. >> >>OK. And how RAID create in Lustre? > > > Lustre itself will currently only do RAID0 (striping) across storage > nodes (OSTs), though we have OST RAID1 support underway for a later release. > > Lustre uses regular block devices to hold the back-end filesystems, so > it can use any kind of device that Linux supports (e.g. LVM/DM, MD RAID, > FC/SCSI/SATA hardware RAID, etc). If the back-end storage is redundant > then it can handle a disk failure internally. > If that storage is also > capable of connecting to multiple nodes (e.g. FC) then Lustre can be > configured to handle node failures (though at the current time this > involves some effort on the part of the implementer).I.e. in Lustre not present internal failure repair mechanisms? :(
--//IivP0gvsAy3Can Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Apr 14, 2005 11:06 -0500, Paul Jimenez wrote:> Is it possible to use Lustre in conjunction with GNBD to obtain=20 > mirrored-failover capability?Several people have reported trying this, but I am not aware of conclusive results either way. One issue is whether GNBD preserves the ordering of IO requests to keep the ext3 journaling happy and the filesystem consistent in case of a crash. Cheers, Andreas -- Andreas Dilger Principal Software Engineer Cluster File Systems, Inc. --//IivP0gvsAy3Can Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.3 (GNU/Linux) iD8DBQFCXtVHpIg59Q01vtYRAuojAJ4h2QcAW+CY0kLDIzC2BgtOVVelBACg2Afp Mzy6hfsqw0fJiLf4GGqU+xE=wOO9 -----END PGP SIGNATURE----- --//IivP0gvsAy3Can--
Is it possible to use Lustre in conjunction with GNBD to obtain mirrored-failover capability? --pj Andreas Dilger wrote:>On Apr 13, 2005 13:52 +0400, Andrew Velikoredchanin wrote: > > >>Andreas Dilger wrote: >> >> >>>On Apr 12, 2005 22:42 +0400, Andrew Velikoredchanin wrote: >>> >>> >>>>I want install Lustre FS in ouer Linux claster. But we need duplicate >>>>data. I read about replication on Lustre, but I do not undestend what >>>>need doing for this. Please, answer - what need for it? >>>> >>>> >>>Lustre does not currently support replication - this is a feature planned >>>for the future. >>> >>>Currently, systems run with some sort of RAID for the back-end storage to >>>provide redundancy in case of disk failure. If the RAID can be attached >>>to multiple nodes (e.g. FC, or multiport SCSI) then it is also possible >>>to provide server failover in case of node failure. >>> >>> >>OK. And how RAID create in Lustre? >> >> > >Lustre itself will currently only do RAID0 (striping) across storage >nodes (OSTs), though we have OST RAID1 support underway for a later release. > >Lustre uses regular block devices to hold the back-end filesystems, so >it can use any kind of device that Linux supports (e.g. LVM/DM, MD RAID, >FC/SCSI/SATA hardware RAID, etc). If the back-end storage is redundant >then it can handle a disk failure internally. If that storage is also >capable of connecting to multiple nodes (e.g. FC) then Lustre can be >configured to handle node failures (though at the current time this >involves some effort on the part of the implementer). > >Cheers, Andreas >-- >Andreas Dilger >Principal Software Engineer >Cluster File Systems, Inc. > >_______________________________________________ >Lustre-discuss mailing list >Lustre-discuss@lists.clusterfs.com >https://lists.clusterfs.com/mailman/listinfo/lustre-discuss > >