First I allocated 300G for MDS. Then I find MDS space is not enough. So I need to add some space to the MDS. Please tell me how to add space for MDS? Thank you!
Look on the lustre wiki for filesystem backup. The procedure is pretty straightforward. I''ve done it four or so times myself and had no problems. On Jun 12, 2008, at 7:30 AM, Johnlya wrote:> First I allocated 300G for MDS. Then I find MDS space is not enough. > So I need to add some space to the MDS. Please tell me how to add > space for MDS? Thank you! > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss at lists.lustre.org > http://lists.lustre.org/mailman/listinfo/lustre-discuss
On Jun 13, 2008, at 12:31 AM1, Aaron Knister <aaron.knis... at gmail.com> wrote:> Look on the lustre wiki for filesystem backup. The procedure is pretty ? > straightforward. I''ve done it four or so times myself and had no ? > problems. > > On Jun 12, 2008, at 7:30 AM,Johnlyawrote: > > > First I allocated 300G for MDS. Then I find MDS space is not enough. > > So I need to add some space to the MDS. Please tell me how to add > > space for MDS? Thank you! > > _______________________________________________ > > Lustre-discuss mailing list > > Lustre-disc... at lists.lustre.org > >http://lists.lustre.org/mailman/listinfo/lustre-discuss > > _______________________________________________ > Lustre-discuss mailing list > Lustre-disc... at lists.lustre.orghttp://lists.lustre.org/mailman/listinfo/lustre-discussHi, Aaron. Do you mean the information "15.1 Lustre Backups:Lustre provides filesystem backups at several levels. " ? I mean that I don''t format(mkfs.lustre --fsname=test --mdt --mgs -- reformat /dev/sda1) the MDS when I add space of MDS. Please tell me how to do it more detailed. Thank you !
On Jun 13, 2008, at 12:31 AM, Johnlya <john... at gmail.com> wrote:> On Jun 13, 2008, at 12:31 AM1, Aaron Knister <aaron.knis... at gmail.com> > wrote: > > > > > > > Look on the lustre wiki for filesystem backup. The procedure is pretty ? > > straightforward. I''ve done it four or so times myself and had no ? > > problems. > > > On Jun 12, 2008, at 7:30 AM,Johnlyawrote: > > > > First I allocated 300G for MDS. Then I find MDS space is not enough. > > > So I need to add some space to the MDS. Please tell me how to add > > > space for MDS? Thank you! > > > _______________________________________________ > > > Lustre-discuss mailing list > > > Lustre-disc... at lists.lustre.org > > >http://lists.lustre.org/mailman/listinfo/lustre-discuss > > > _______________________________________________ > > Lustre-discuss mailing list > > Lustre-disc... at lists.lustre.orghttp://lists.lustre.org/mailman/listinfo/lustre-discuss > > Hi, Aaron. Do you mean the information "15.1 Lustre Backups:Lustre > provides filesystem backups at several levels. " ? > > I mean that I don''t format(mkfs.lustre --fsname=test --mdt --mgs -- > reformat /dev/sda1) the MDS when I add space of MDS. > > Please tell me how to do it more detailed. Thank you ! > _______________________________________________ > Lustre-discuss mailing list > Lustre-disc... at lists.lustre.orghttp://lists.lustre.org/mailman/listinfo/lustre-discuss- >below step is correct: backup MDS: mount -t ldiskfs /dev/sda1 /mnt/test/mdt/ getfattr -R -d -m ''.*'' -P . > ea.bak tar czvf /home/test/mdtbackup.tgz ./ umount /mnt/test/mdt/ restore MDS: mkfs.lustre --fsname=test --mdt --mgs --reformat /dev/hda4 mke2fs -j -J size=400 -I 512 -i 4096 /dev/hda4 tune2fs -O dir_index /dev/hda4 mount -t ldiskfs /dev/hda4 /mnt/test/mdt/ tar xzvpf /home/test/mdtbackup.tgz setfattr --restore=ea.bak rm OBJECTS/* CATALOGS umount /mnt/test/mdt/ mount.lustre /dev/hda4 /mnt/test/mdt I tested it and it can work. But I don''t do it by its method. I want to add space like adding OST. Thank you!
On Fri, 2008-06-13 at 04:35 -0700, Johnlya wrote:> I tested it and it can work. But I don''t do it by its method. I want > to add space like adding OST.That is currently not possible. The only method we support for MDT expansion is the "backup/recreate MDT/restore" process. Cheers, b. -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: This is a digitally signed message part Url : http://lists.lustre.org/pipermail/lustre-discuss/attachments/20080616/5420c19f/attachment.bin
On Mon, 2008-06-16 at 21:38, "Brian J. Murrell" <Brian.Murr... at Sun.COM> wrote:> On Fri, 2008-06-13 at 04:35 -0700, Johnlya wrote: > > I tested it and it can work. But I don''t do it by its method. I want > > to add space like adding OST. > > That is currently not possible. ?The only method we support for MDT > expansion is the "backup/recreate MDT/restore" process. > > Cheers, > b. > > ?signature.asc > 1Kdownload > > _______________________________________________ > Lustre-discuss mailing list > Lustre-disc... at lists.lustre.orghttp://lists.lustre.org/mailman/listinfo/lustre-discussThanks a lot!
Does Lustre support memory mapped I/O and direct I/O? I am trying to run Nastran using Lustre but it always reported failure to create a directory. Since Nastran does a lot of memory mapped I/O, I was wondering if it was the cause. I guess a good question to ask is that does Lustre support all POSIX file system operations? Thanks a lot. Eric
Huang, Eric writes: Hello, > > Does Lustre support memory mapped I/O and direct I/O? yes, it supports both. Can you run your application under strace to see how exactly it fails to create a directory? > > I am trying to run Nastran using Lustre but it always reported failure > to create a directory. Since Nastran does a lot of memory mapped I/O, I > was wondering if it was the cause. > > I guess a good question to ask is that does Lustre support all POSIX > file system operations? > > Thanks a lot. > > Eric Nikita.
On Jun 17, 2008 22:39 -0700, Huang, Eric wrote:> Does Lustre support memory mapped I/O and direct I/O?Yes, it does support both of these.> I am trying to run Nastran using Lustre but it always reported failure > to create a directory. Since Nastran does a lot of memory mapped I/O, I > was wondering if it was the cause.??? I''m not sure how mmap and direct I/O relate to creating a directory?> I guess a good question to ask is that does Lustre support all POSIX > file system operations?Yes. Cheers, Andreas -- Andreas Dilger Sr. Staff Engineer, Lustre Group Sun Microsystems of Canada, Inc.
On Jun 24, 2008 17:35 -0700, Huang, Eric wrote:> Got another question, does Lustre provide a different user level library > to access the file system? I mean, when the user application does a > "read" or "write", is it using the regular Linux library to conduct > these system calls? Or does Lustre provide its own, such that these > "read" and "write" calls are not typical VFS calls anymore? My > understanding is that Lustre has its own user-level library that > intercepts these system calls and send commands across Lnet to get/write > data. Only the meta-data related calls (such as INODE allocation, > deletion, etc.) are actually passed to VFS. Am I right?No, Lustre is a regular filesystem that is POSIX compliant. There IS a userspace Library for Lustre, but that is used only for a very limited class of systems, normally the Catamount clients on the Cray XT3, and if non-Linux clients are using FUSE to access Lustre (e.g. OSX, FreeBSD).> -----Original Message----- > From: Andreas.Dilger at sun.com [mailto:Andreas.Dilger at sun.com] On Behalf > Of Andreas Dilger > Sent: Monday, June 23, 2008 3:56 PM > To: Huang, Eric > Cc: Nikita.Danilov at sun.com > Subject: Re: [Lustre-discuss] Lustre and memory-mapped I/O > > On Jun 19, 2008 10:18 -0700, Huang, Eric wrote: > > Here is the error: > > > > *** USER FATAL MESSAGE 1250 (BIOCRT) > > STATUS = 38, FILX = 3, LOGNAME = MASTER , NSBUF3 > > 65536 > > FILE = /lustre/client/4way/xx0cmd0.T18856_24.t7.MASTER > > ERROR MESSAGE IS -- > > Function not implemented (errno = 38) > > *** SYSTEM FATAL MESSAGE 446 (DBCRT) > > ERROR ATTEMPTING TO CREATE A FILE (LOGICAL= MASTER ) FOR DBSET > > MASTER WITH READ AND WRITE PRIVILEGES. > > USER ACTION: CHECK THE AVAILABLE/ALLOCATED SPACE AND THE > PRIVILEGES > > GRANTED THIS ACCOUNT. > > > > USER INFORMATION: THIS ERROR OCCURS WHEN MORE THAN ONE JOB HAS > WRITE > > PRIVILEGE AT THE SAME TIME. > > ONLY ONE JOB CAN HAVE WRITE PRIVILEGE. > > 0FATAL ERROR > > 1 * * * END OF JOB * * * > > > > > > I think it may have to do with the Lustre lock implementation. The > > directory /lustre/client/4way is a directory with file-striping set to > 4. > > Is this application trying to use flock? Then the client needs to > be mounted with flock support, either "-o flock" to enable distributed > flock (lower performance), or "-o localflock" (for local node flock > only, > if application only needs flock on a single node). > > Cheers, Andreas > -- > Andreas Dilger > Sr. Staff Engineer, Lustre Group > Sun Microsystems of Canada, Inc.Cheers, Andreas -- Andreas Dilger Sr. Staff Engineer, Lustre Group Sun Microsystems of Canada, Inc.