[LU-60] mkfs.lustre's --mgsnode=nid:nid syntax not working

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.1.0, Lustre 1.8.6
• Type: Bug
• Priority: Minor
• Reporter: Brian Murrell (Inactive)
• Assignee: WC Triage
• Resolution: Incomplete
• Severity: 3

Description

Given the following node mapping:

192.168.122.147	oss1
192.168.122.160	oss2
192.168.122.89	mds1
192.168.122.115	mds2

When I format an OST with the following syntax:

oss1# mkfs.lustre --ost --reformat --mgsnode=mds1:mds2 --failnode=oss2 /dev/vda

and then try to mount it, the mount fails with dmesg reporting:

Lustre: 2876:0:(client.c:1476:ptlrpc_expire_one_request()) @@@ Request x1359778535178425 sent from MGC192.168.122.89@tcp to NID 192.168.122.115@tcp 5s ago has timed out (5s prior to deadline).
  req@ffff8100174bcc00 x1359778535178425/t0 o250->MGS@MGC192.168.122.89@tcp_0:26/25 lens 368/584 e 0 to 1 dl 1296831448 ref 1 fl Rpc:N/0/0 rc 0/0
LustreError: 6413:0:(obd_mount.c:1097:server_start_targets()) Required registration failed for lustre-OSTffff: -4
LustreError: 6413:0:(obd_mount.c:1655:server_fill_super()) Unable to start targets: -4
LustreError: 6413:0:(obd_mount.c:1438:server_put_super()) no obd lustre-OSTffff
LustreError: 6413:0:(obd_mount.c:147:server_deregister_mount()) lustre-OSTffff not registered

However if I format with the following syntax:

oss1# mkfs.lustre --ost --reformat --mgsnode=mds1 --mgsnode=mds2 --failnode=oss2 /dev/vda

mounting succeeds.

Just for good measure I tried a , instead of a : between the NIDs and the result is the same as for the :.

Comments

[Cliff White (Inactive)]

Per the 1.8 manual, section 4.4 - the comma ',' indicates multiple NIDS for a single node, odd that the result would be the same. Could you attach a tune2fs printout for the ':' case?

[Brian Murrell (Inactive)]

After the format, prior to the initial mount (i.e. registration):

# tunefs.lustre --print /dev/vda
checking for existing Lustre data: found CONFIGS/mountdata
Reading CONFIGS/mountdata

   Read previous values:
Target:     lustre-OSTffff
Index:      unassigned
Lustre FS:  lustre
Mount type: ldiskfs
Flags:      0x72
              (OST needs_index first_time update )
Persistent mount opts: errors=remount-ro,extents,mballoc
Parameters: mgsnode=192.168.122.89@tcp:192.168.122.115@tcp failover.node=192.168.122.160@tcp


   Permanent disk data:
Target:     lustre-OSTffff
Index:      unassigned
Lustre FS:  lustre
Mount type: ldiskfs
Flags:      0x72
              (OST needs_index first_time update )
Persistent mount opts: errors=remount-ro,extents,mballoc
Parameters: mgsnode=192.168.122.89@tcp:192.168.122.115@tcp failover.node=192.168.122.160@tcp

exiting before disk write.

And just for good measure (and comparison purposes) here is the tunefs.lustre --print output after formatting with two --mgsnode= specifiers:

# tunefs.lustre --print /dev/vda
checking for existing Lustre data: found CONFIGS/mountdata
Reading CONFIGS/mountdata

   Read previous values:
Target:     lustre-OSTffff
Index:      unassigned
Lustre FS:  lustre
Mount type: ldiskfs
Flags:      0x72
              (OST needs_index first_time update )
Persistent mount opts: errors=remount-ro,extents,mballoc
Parameters: mgsnode=192.168.122.89@tcp mgsnode=192.168.122.115@tcp failover.node=192.168.122.160@tcp


   Permanent disk data:
Target:     lustre-OSTffff
Index:      unassigned
Lustre FS:  lustre
Mount type: ldiskfs
Flags:      0x72
              (OST needs_index first_time update )
Persistent mount opts: errors=remount-ro,extents,mballoc
Parameters: mgsnode=192.168.122.89@tcp mgsnode=192.168.122.115@tcp failover.node=192.168.122.160@tcp

exiting before disk write.

[Brian Murrell (Inactive)]

I was also able to reproduce this on master at 0091fc7, which is admittedly quite old, but I suspect this has not been addressed yet.

[Build Master (Inactive)]

Integrated in hydra-agent » x86_64,el6 #84
Work around lu-60

[Build Master (Inactive)]

Integrated in hydra-agent » x86_64,el5 #84
Work around lu-60

[Build Master (Inactive)]

Integrated in chroma » x86_64,el6 #3
Work around lu-60 (Revision 475e12d024964ec9a5801d601cd6ca9d230aaa33)

Result = SUCCESS

[Andreas Dilger]

Closing this old ticket.