Patrice Bouchand
2006-Jul-05 05:29 UTC
[Lustre-discuss] How to set lov.stripe.size with tunefs.lustre
Hello, I''m working with lustre 1.5.91, and I''m trying to set the stripe size with tunefs.lustre. /dev/sdb2 is an MDT/MGT. Here''s what I do: The filesystem is formated and online. The stripe size is the default: [root@ns3 dev_16]#cat /proc/fs/lustre/lov/fs1-mdtlov/stripesize 1048576 I stop the filesystem and I set the stripe size to 4M using tunefs.lustre: [root@ns3 dev_16]# tunefs.lustre --param lov.stripe.size=4194304 /dev/sdb2 checking for existing Lustre data found Lustre data Reading CONFIGS/mountdata Read previous values: Target: fs1-MDT0000 Index: 0 UUID: Lustre FS: fs1 Mount type: ldiskfs Flags: 0x5 (MDT MGS ) Persistent mount opts: errors=remount-ro,iopen_nopriv,user_xattr,user_xattr,acl Parameters: Permanent disk data: Target: fs1-MDT0000 Index: 0 UUID: Lustre FS: fs1 Mount type: ldiskfs Flags: 0x5 (MDT MGS ) Persistent mount opts: errors=remount-ro,iopen_nopriv,user_xattr,user_xattr,acl Parameters: lov.stripe.size=4194304 Writing CONFIGS/mountdata I start again the filesystem and the stripe size is still 1M. [root@ns3 dev_16]# cat /proc/fs/lustre/lov/fs1-mdtlov/stripesize 1048576 Any idea about what I''m doing wrong ? Patrice Bouchand -------------- next part -------------- An HTML attachment was scrubbed... URL: http://mail.clusterfs.com/pipermail/lustre-discuss/attachments/20060705/93d59151/attachment.html
Niklas Edmundsson
2006-Jul-05 07:12 UTC
[Lustre-discuss] How to set lov.stripe.size with tunefs.lustre
On Wed, 5 Jul 2006, Patrice Bouchand wrote:> I stop the filesystem and I set the stripe size to 4M using tunefs.lustre: > > [root@ns3 dev_16]# tunefs.lustre --param lov.stripe.size=4194304 /dev/sdb2 > checking for existing Lustre data > found Lustre data > Reading CONFIGS/mountdata> Any idea about what I''m doing wrong ?Read https://mail.clusterfs.com/wikis/lustre/MountConf section "Changing parameters" ... /Nikke -- -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Niklas Edmundsson, Admin @ {acc,hpc2n}.umu.se | nikke@hpc2n.umu.se --------------------------------------------------------------------------- Life''s not a spectacle or a feast, it''s a predicament =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Patrice Bouchand
2006-Jul-05 07:38 UTC
[Lustre-discuss] How to set lov.stripe.size with tunefs.lustre
> Read https://mail.clusterfs.com/wikis/lustre/MountConf section > "Changing parameters" ...What I need to do is to change stripe size or such parameters on an offline and already formated MGS. I thought it was tunefs.lustre job ? -- Patrice BOUCHAND patrice.bouchand@ext.bull.net Bull Echirolles B1-430 tel : +4 76 29 75 23
Peter J. Braam
2006-Jul-05 09:01 UTC
[Lustre-discuss] How to set lov.stripe.size with tunefs.lustre
Hi Tunefs mostly tunes the backends. What you want to use here is the "lfs" utility - the description in the manual is not great, but it should give you enough information to change the stripe size and stripe count if you wish. - Peter - -----Original Message----- From: lustre-discuss-bounces@clusterfs.com [mailto:lustre-discuss-bounces@clusterfs.com] On Behalf Of Patrice Bouchand Sent: Wednesday, July 05, 2006 7:38 AM To: lustre-discuss@clusterfs.com Cc: Niklas Edmundsson Subject: Re: [Lustre-discuss] How to set lov.stripe.size with tunefs.lustre> Read https://mail.clusterfs.com/wikis/lustre/MountConf section > "Changing parameters" ...What I need to do is to change stripe size or such parameters on an offline and already formated MGS. I thought it was tunefs.lustre job ? -- Patrice BOUCHAND patrice.bouchand@ext.bull.net Bull Echirolles B1-430 tel : +4 76 29 75 23 _______________________________________________ Lustre-discuss mailing list Lustre-discuss@clusterfs.com https://mail.clusterfs.com/mailman/listinfo/lustre-discuss
green@clusterfs.com
2006-Jul-05 09:12 UTC
[Lustre-discuss] How to set lov.stripe.size with tunefs.lustre
Hello! On Wed, Jul 05, 2006 at 11:01:29AM -0400, Peter J. Braam wrote:> Hi > > Tunefs mostly tunes the backends. What you want to use here is the > "lfs" utility - the description in the manual is not great, but it > should give you enough information to change the stripe size and stripe > count if you wish.Also there is KB entry on striping. https://bugzilla.lustre.org/show_bug.cgi?id=9357 It was written for 1.4.x, but lfs setstripe description is still correct. Bye, Oleg
Patrice Bouchand
2006-Jul-05 09:48 UTC
[Lustre-discuss] How to set lov.stripe.size with tunefs.lustre
In fact, stripping is not really my problem. What I need to do is something like lconf --writeconf with an updated XML, but with lustre 1.6, and I need to do this on an offline filesystem. We (Bull) are updating our administration tools to work with lustre 1.6, and it seems that tunefs.lustre is the closest way to do such a thing. So, if tunefs.lustre --param lov.stripe.size= does not aim to change the stripe size of an offline filesystem, what should it do ? Le mercredi 05 juillet 2006 ? 18:13 +0300, green@clusterfs.com a ?crit :> Hello! > > On Wed, Jul 05, 2006 at 11:01:29AM -0400, Peter J. Braam wrote: > > Hi > > > > Tunefs mostly tunes the backends. What you want to use here is the > > "lfs" utility - the description in the manual is not great, but it > > should give you enough information to change the stripe size and stripe > > count if you wish. > > Also there is KB entry on striping. > https://bugzilla.lustre.org/show_bug.cgi?id=9357 > It was written for 1.4.x, but lfs setstripe description is still correct. > > Bye, > Oleg > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss@clusterfs.com > https://mail.clusterfs.com/mailman/listinfo/lustre-discuss
Nathaniel Rutman
2006-Jul-05 19:05 UTC
[Lustre-discuss] How to set lov.stripe.size with tunefs.lustre
Please see https://mail.clusterfs.com/wikis/lustre/MountConf#head-7284705c97385ac66da935aff8cabc58faa0f0a1 On a running mdt, do something like mgs> lctl conf_param testfs-MDT0000 lov.stripe.size=4194304 (The tunefs.lustre interface is not yet able to modify logs, so although you can change the parameter, only the parameters set when the server is _first mounted_ get used.) Patrice Bouchand wrote:> Hello, > > I''m working with lustre 1.5.91, and I''m trying to set the stripe size > with tunefs.lustre. /dev/sdb2 is an MDT/MGT. Here''s what I do: > > The filesystem is formated and online. The stripe size is the default: > > [root@ns3 dev_16]#cat /proc/fs/lustre/lov/fs1-mdtlov/stripesize > > 1048576 > > I stop the filesystem and I set the stripe size to 4M using tunefs.lustre: > > [root@ns3 dev_16]# tunefs.lustre --param lov.stripe.size=4194304 /dev/sdb2 > > checking for existing Lustre data > > found Lustre data > > Reading CONFIGS/mountdata > > Read previous values: > > Target: fs1-MDT0000 > > Index: 0 > > UUID: > > Lustre FS: fs1 > > Mount type: ldiskfs > > Flags: 0x5 > > (MDT MGS ) > > Persistent mount opts: > errors=remount-ro,iopen_nopriv,user_xattr,user_xattr,acl > > Parameters: > > Permanent disk data: > > Target: fs1-MDT0000 > > Index: 0 > > UUID: > > Lustre FS: fs1 > > Mount type: ldiskfs > > Flags: 0x5 > > (MDT MGS ) > > Persistent mount opts: > errors=remount-ro,iopen_nopriv,user_xattr,user_xattr,acl > > Parameters: lov.stripe.size=4194304 > > Writing CONFIGS/mountdata > > I start again the filesystem and the stripe size is still 1M. > > [root@ns3 dev_16]# cat /proc/fs/lustre/lov/fs1-mdtlov/stripesize > > 1048576 > > Any idea about what I''m doing wrong ? > > Patrice Bouchand > >------------------------------------------------------------------------ > >_______________________________________________ >Lustre-discuss mailing list >Lustre-discuss@clusterfs.com >https://mail.clusterfs.com/mailman/listinfo/lustre-discuss > >
Nathaniel Rutman
2006-Jul-05 19:15 UTC
[Lustre-discuss] How to set lov.stripe.size with tunefs.lustre
Patrice, you are entirely correct, and that is a goal of tunefs.lustre. However, it is likely that this method will not be implemented in the first release of 1.6. You can currently change the parameters while the servers are online using the method described in the wiki, or the other knowledge base articles. Patrice Bouchand wrote:>In fact, stripping is not really my problem. > What I need to do is something like lconf --writeconf with an updated >XML, but with lustre 1.6, and I need to do this on an offline >filesystem. We (Bull) are updating our administration tools to work with >lustre 1.6, and it seems that tunefs.lustre is the closest way to do >such a thing. So, if tunefs.lustre --param lov.stripe.size= does not aim >to change the stripe size of an offline filesystem, what should it do ? > > >Le mercredi 05 juillet 2006 ? 18:13 +0300, green@clusterfs.com a ?crit : > > >>Hello! >> >>On Wed, Jul 05, 2006 at 11:01:29AM -0400, Peter J. Braam wrote: >> >> >>>Hi >>> >>>Tunefs mostly tunes the backends. What you want to use here is the >>>"lfs" utility - the description in the manual is not great, but it >>>should give you enough information to change the stripe size and stripe >>>count if you wish. >>> >>> >>Also there is KB entry on striping. >>https://bugzilla.lustre.org/show_bug.cgi?id=9357 >>It was written for 1.4.x, but lfs setstripe description is still correct. >> >>Bye, >> Oleg >>_______________________________________________ >>Lustre-discuss mailing list >>Lustre-discuss@clusterfs.com >>https://mail.clusterfs.com/mailman/listinfo/lustre-discuss >> >> > >_______________________________________________ >Lustre-discuss mailing list >Lustre-discuss@clusterfs.com >https://mail.clusterfs.com/mailman/listinfo/lustre-discuss > >
Patrice Bouchand
2006-Jul-05 23:40 UTC
[Lustre-discuss] How to set lov.stripe.size with tunefs.lustre
Ok, thanks a lot ! Patrice> Patrice, you are entirely correct, and that is a goal of tunefs.lustre. > However, it is likely that this method will not > be implemented in the first release of 1.6. > You can currently change the parameters while the servers are online > using the method described in the wiki, > or the other knowledge base articles.