Hi everyone, I am trying to upgrade to Lustre 1.6.6 and having a problem. Our system is using one OSS and one MGS, the number of OST is 24. So far, it seems I have succeeded to mkfs.lustre to the 24 OST(from /dev/sda, sdb.......sdx) and succeeded to mount.lustre those to /mnt/ost0 ost1.....ost23. However, MGS doesn''t seem to be working properly. After mount.lustre /dev/VolGroup00/mdt /mnt/mdt/, checking /proc/fs/lustre/mds/user-MDT0000/recovery_status returns "status: INACTIVE". In face, the first time I have mounted to the 24 OST, it also returned "status: INACTIVE". But after activated all 24 OSTs using "lctl activate num" command, all of those recovery_status became "COMPLETE". However, even if I did the same thing for MGS, the recovery_status never become "COMPLETE". Would someone kindly please help me to figure this problem out? or any advise to solve this out would be really appreciated. Thanks in advance for your time, Shigeru -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.lustre.org/pipermail/lustre-discuss/attachments/20090221/a3cdfcff/attachment.html
On Feb 21, 2009 01:24 +0900, Shigeru Sugimoto wrote:> I am trying to upgrade to Lustre 1.6.6 and having a problem. Our system > is using one OSS and one MGS, the number of OST is 24. > > So far, it seems I have succeeded to mkfs.lustre to the 24 OST(from > /dev/sda, sdb.......sdx) and succeeded to mount.lustre those to /mnt/ost0 > ost1.....ost23.You will have severe memory problems under load with this many OSTs. Is there a reason not to make larger OSTs (up to 8TB is supported).> However, MGS doesn''t seem to be working properly. > After mount.lustre /dev/VolGroup00/mdt /mnt/mdt/, checking > /proc/fs/lustre/mds/user-MDT0000/recovery_status > returns "status: INACTIVE".This is not harmful, it just means recovery was never run on this OST, and that is normal if it was shut down cleanly. Cheers, Andreas -- Andreas Dilger Sr. Staff Engineer, Lustre Group Sun Microsystems of Canada, Inc.
Andreas, Thanks for your quick reply, it really helps me. 2009/2/21 Andreas Dilger adilger at sun.com> > You will have severe memory problems under load with this many OSTs. > Is there a reason not to make larger OSTs (up to 8TB is supported).Sorry, I am still a junior lustre user and don''t know how to configure OST size. Would you please kindly tell me how to configure the OST size? or just give me the pointer to those info is greatly appreciated. I am looking at the 1.6 version of Lustre Operations Manual but can''t find anything related to this info. By looking at /proc/fs/lustre/obdfilter/user-OST0000/kbytestotal, it seems my OST size is currently configured to around 1.3TB. Thanks in advance for your kindness. Shigeru -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.lustre.org/pipermail/lustre-discuss/attachments/20090223/1b2eb78a/attachment-0001.html