Chad Kerner
2008-Aug-06 12:16 UTC
[Lustre-discuss] question with mount.lustre truncating device name
Hello, We are trying to get a 1.6.5.1 lustre system up and running. We are using udev to map out our fiber channel devices. We have the following disk /dev/sdb with the scsi_id below. [root at abe-io07 ~]# ls_fc_luns HBA 210000e08b0db5cb host1 210000e08b0db5cb 0x200200a0b80f4498 1:0:0:0 sdb 3600a0b80000f449700000000487dc46a [root at abe-io07 ~] The udev rule looks like this: [root at abe-io07 ~]# grep "ost_l_01" /etc/udev/rules.d/20-local.rules BUS="scsi", SYSFS{vendor}="IBM", SYSFS{model}="1742", PROGRAM="/sbin/scsi_id -g -u -s /block/%k ", RESULT="3600a0b80000f449700000000487dc46a", NAME="disk/fc/fastt6e_0" SYMLINK="disk/lustre/ost_l_01" [root at abe-io07 ~]# The device is there properly named: [root at abe-io07 ~]# ll /dev/sdb brw------- 1 root root 8, 16 Aug 5 07:10 /dev/sdb [root at abe-io07 ~]# ll /dev/disk/fc/fastt6e_0 brw-rw---- 1 root disk 8, 16 Aug 5 12:10 /dev/disk/fc/fastt6e_0 [root at abe-io07 ~]# ll /dev/disk/lustre/ost_l_01 lrwxrwxrwx 1 root root 15 Aug 5 12:12 /dev/disk/lustre/ost_l_01 -> ../fc/fastt6e_0 We did a mkfs.lustre on it and it said it was successful(we used the /dev/disk/lustre/ost_l_01 device for the format). But when we try to mount the device we get the following error: [root at abe-io07 ~]# mount -v -t lustre /dev/disk/lustre/ost_l_01 /lustre/local/ost_l_01 arg[0] = /sbin/mount.lustre arg[1] = -v arg[2] = -o arg[3] = rw arg[4] = /dev/disk/lustre/ost_l_01 arg[5] = /lustre/local/ost_l_01 source = /dev/disk/lustre/ost_l_01 (/dev/disk/lustre/ost_l_01), target = /lustre/local/ost_l_01 options = rw mounting device /dev/disk/lustre/ost_l_01 at /lustre/local/ost_l_01, flags=0 options=device=/dev/disk/lustre/ost_l_01 warning: No such file or directory, device /dev/disk/fc/fastt stat failed /sbin/mount.lustre: unable to set tunables for /dev/disk/lustre/ost_l_01 (may cause reduced IO performance)mount.lustre: mount /dev/disk/lustre/ost_l_01 at /lustre/local/ost_l_01 failed: Input/output error Is the MGS running? [root at abe-io07 ~] The warning: ... the device name is getting truncated. It looks like from mount_lustre.c that any device name that contains a numeric digit in the device name will get truncated. /* The name of the device say ''X'' specified in /dev/X may not * match any entry under /sys/block/. In that case we need to * match the major/minor number to find the entry under * sys/block corresponding to /dev/X */ dev = real_path + src_len - 1; while (dev > real_path) { if (isdigit(*dev)) *dev = 0; dev--; } rc = stat(dev, &stat_buf); if (rc) { if (verbose) fprintf(stderr, "warning: %s, device %s stat failed\n", strerror(errno), dev); return rc; } Was looking for some suggestions on getting around this issue. Thanks, Chad
Brian J. Murrell
2008-Aug-06 16:37 UTC
[Lustre-discuss] question with mount.lustre truncating device name
On Wed, 2008-08-06 at 07:16 -0500, Chad Kerner wrote:> Hello, > > We are trying to get a 1.6.5.1 lustre system up and running. We > are using udev to map out our fiber channel devices....> It looks like from mount_lustre.c that any device name that contains a > numeric digit in the device name will get truncated.Indeed. You can change the NAME="disk/fc/fastt6e_0" in your udev rule to match the device naming convention that is being assumed by mount.lustre and that''s that the only digits in the last path component are postfixed and identify partitions. If you don''t install your own udev rule, what does the kernel name that device by default? Does it match the assumption that mount.lustre is making? 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/20080806/dcf3ec31/attachment.bin