[LU-3830] mount fails on targets configured with multiple mgs nids

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.4.0
• Type: Bug
• Priority: Major
• Reporter: Gregoire Pichon
• Assignee: WC Triage
• Resolution: Duplicate
• Severity: 3
• Epic: mount

Description

When a target is configured with multiple nids, the mount fails due to invalid mount options.

# mkfs.lustre --mgs --fsname=localfs /dev/ram0
# mkfs.lustre --mdt --fsname=localfs --mgsnode=30.1.0.95@o2ib0 --index=0 /dev/ram1
# mkfs.lustre --ost --fsname=localfs --mgsnode=30.1.0.95@o2ib0 --index=0 /dev/ram2
# mkfs.lustre --ost --fsname=localfs --mgsnode=30.1.0.95@o2ib0,30.2.0.95@o2ib1 --index=1 /dev/ram3

# mount.lustre /dev/ram0 /mnt/localfs/mgt
# mount.lustre /dev/ram1 /mnt/localfs/mdt
# mount.lustre /dev/ram2 /mnt/localfs/ost0
# mount.lustre /dev/ram3 /mnt/localfs/ost1
mount.lustre: mount /dev/ram3 at /mnt/localfs/ost1 failed: Invalid argument
This may have multiple causes.
Are the mount options correct?
Check the syslog for more info.

# mount.lustre /dev/ram3 /mnt/localfs/ost1 -v
arg[0] = /sbin/mount.lustre.orig
arg[1] = -v
arg[2] = /dev/ram3
arg[3] = /mnt/localfs/ost1
source = /dev/ram3 (/dev/ram3), target = /mnt/localfs/ost1
options = 
checking for existing Lustre data: found
Reading CONFIGS/mountdata
mounting device /dev/ram3 at /mnt/localfs/ost1, flags=0x1000000 options=osd=osd-ldiskfs,errors=remount-ro,mgsnode=30.1.0.95@o2ib,30.2.0.95@o2ib1,virgin,param=mgsnode=30.1.0.95@o2ib,30.2.0.95@o2ib1,svname=localfs-OST0001,device=/dev/ram3
mount.lustre: mount /dev/ram3 at /mnt/localfs/ost1 failed: Invalid argument retries left: 0
mount.lustre: mount /dev/ram3 at /mnt/localfs/ost1 failed: Invalid argument
This may have multiple causes.
Are the mount options correct?
Check the syslog for more info.

# tail -n 6 /var/log/syslog
1377250637 2013 Aug 23 11:37:17 mo95 kern err kernel LDISKFS-fs (ram3): Unrecognized mount option "30.2.0.95@o2ib1" or missing value
1377250637 2013 Aug 23 11:37:17 mo95 kern err kernel LustreError: 32322:0:(osd_handler.c:5349:osd_mount()) localfs-OST0001-osd: can't mount /dev/ram3: -22
1377250637 2013 Aug 23 11:37:17 mo95 kern err kernel LustreError: 32322:0:(obd_config.c:572:class_setup()) setup localfs-OST0001-osd failed (-22)
1377250637 2013 Aug 23 11:37:17 mo95 kern err kernel LustreError: 32322:0:(obd_mount.c:201:lustre_start_simple()) localfs-OST0001-osd setup error -22
1377250637 2013 Aug 23 11:37:17 mo95 kern err kernel LustreError: 32322:0:(obd_mount_server.c:1665:server_fill_super()) Unable to start osd on /dev/ram3: -22
1377250637 2013 Aug 23 11:37:17 mo95 kern err kernel LustreError: 32322:0:(obd_mount.c:1267:lustre_fill_super()) Unable to mount  (-22)

It seems that mountdata options are now passed to mount.lustre as additional options.

Unfortunately comma is used as a separator of mount options and multiple mgs nids are also separated by a comma.

This lead to have the second mgs nid "30.2.0.95@o2ib1" to be processed as a mount option, which fails

Issue Links

• duplicates: LU-3445 Specifying multiple networks in NIDs ...

Comments

[Niu Yawei (Inactive)]

It should have been fixed in LU-3445.

[Bruno Faccini (Inactive)]

Yes, Niu you have been too fast and I definitely agree! Gregoire, can you give a try to changes/fixes for LU-3445 ??

[Gregoire Pichon]

Thanks for the notice. I missed it when looking for existing Jira tickets.
This is definitively fixed by LU-3445.
Ticket can be closed.

[Bruno Faccini (Inactive)]

Ticket is a duplicate of LU-3445 which is already fixed.