[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[Linux-cluster] RE: values for gnbd multipath dmsetup



Hello,

 

Just thought I’d plug in some more info I’ve also tried testing this with a stock FC4 client with all the Cluster RPMs installed.

 

GFS: fsid=sclients:mygfs.0: jid=17: Trying to acquire journal lock...

GFS: fsid=sclients:mygfs.0: jid=17: Looking at journal...

GFS: fsid=sclients:mygfs.0: jid=17: Done

GFS: fsid=sclients:mygfs.0: jid=18: Trying to acquire journal lock...

GFS: fsid=sclients:mygfs.0: jid=18: Looking at journal...

GFS: fsid=sclients:mygfs.0: jid=18: Done

GFS: fsid=sclients:mygfs.0: jid=19: Trying to acquire journal lock...

GFS: fsid=sclients:mygfs.0: jid=19: Looking at journal...

GFS: fsid=sclients:mygfs.0: jid=19: Done

GFS: fsid=sclients:mygfs.0: jid=20: Trying to acquire journal lock...

GFS: fsid=sclients:mygfs.0: jid=20: Looking at journal...

attempt to access beyond end of device

dm-0: rw=0, want=1146990600, limit=1146990592

GFS: fsid=sclients:mygfs.0: fatal: I/O error

GFS: fsid=sclients:mygfs.0:   block = 143373824

GFS: fsid=sclients:mygfs.0:   function = gfs_dreread

GFS: fsid=sclients:mygfs.0:   file = /usr/src/build/588748-i686/BUILD/xen0/src/gfs/dio.c, line = 576

GFS: fsid=sclients:mygfs.0:   time = 1123268277

GFS: fsid=sclients:mygfs.0: about to withdraw from the cluster

GFS: fsid=sclients:mygfs.0: waiting for outstanding I/O

GFS: fsid=sclients:mygfs.0: telling LM to withdraw

lock_dlm: withdraw abandoned memory

GFS: fsid=sclients:mygfs.0: withdrawn

GFS: fsid=sclients:mygfs.0: jid=20: Failed

GFS: fsid=sclients:mygfs.0: error recovering journal 20: -5

[root 5n@k3bi73 ~]#  

Aug  5 11:57:57 5n k3bi73 kernel: GFS: fsid=sclients:mygfs.0:   time = 1123268277

Aug  5 11:57:57 5n k3bi73 kernel: GFS: fsid=sclients:mygfs.0: about to withdraw from the cluster

Aug  5 11:57:57 5n k3bi73 kernel: GFS: fsid=sclients:mygfs.0: waiting for outstanding I/O

Aug  5 11:57:57 5n k3bi73 kernel: GFS: fsid=sclients:mygfs.0: telling LM to withdraw

Aug  5 11:57:57 5n k3bi73 kernel: lock_dlm: withdraw abandoned memory

Aug  5 11:57:57 5n k3bi73 kernel: GFS: fsid=sclients:mygfs.0: withdrawn

Aug  5 11:57:57 5n k3bi73 kernel: GFS: fsid=sclients:mygfs.0: jid=20: Failed

 

This off a multipathed device which dmsetup status gives the output below:

 

dm-1: 0 1146990592 multipath 1 0 0 2 1 A 0 1 0 251:0 A 0 E 0 1 0 251:4 A 0

dmsetup deps gives

dm-1: 2 dependencies    : (251, 4) (251, 0)

and dmsetup info gives

Name:              dm-1

State:             ACTIVE

Tables present:    LIVE

Open count:        0

Event number:      0

Major, minor:      253, 1

Number of targets: 1

 

 


From: brianu [mailto:brianu silvercash com]
Sent: Friday, August 05, 2005 10:38 AM
To: linux-cluster redhat com
Cc: brianu silvercash com
Subject: RE: values for gnbd multipath dmsetup

 

Hello,

 

Ok I figured out id just try some of the vaules from the previous post without fully understanding them, and multipath appears to be working.

 

dm-1

[size=546 GB][features="0"][hwhandler="0"] \_ round-robin 0 [active][first]

  \_ 0:0:0:0      251:0   [undef ][active]

\_ round-robin 0 [enabled]

  \_ 0:0:0:0      251:4   [undef ][active]

 

 

 

But I stil get an error

 

[root dell-1650-31 ~]# mount -t gfs /dev/mapper/dm-1 /mnt/gfs1

mount: /dev/dm-1: can't read superblock

 

if I do a dmsetup remove dm-1, then mount the individual gnbds all is well, but the purpose of this is to enable some sore of failover which I am told GNBD has the capability of doing.

 

From redhats main site and documentation for gfs 6.1 they state that multipath is not supported in the 6.1 realease however I optained this source from CVS and the main docs for http://sources.redhat.com/cluster/gnbd/  state that multipath is an option. Can someone clarify whether the CVS stabile sources for kernel-2.6.12 is multipath compatable, or am I doing something wrong?

 

Current specs.

SAN -> MSA-1000

 

3 GNBD servers currently using software iSCSI to mount that SAN – will prob go fiber if I can figure this out. ( lets say this cluster is called cluster1)

Using DLM & GNBD

 

1 client for testing separate cluster name lets say “cluster2”

Client mounted the gnbd from one of the servers that is exporting it, the servers are not mounting it, then formatted the device with gfs & created 20 journals size of 32MB each, remounted the device and verified write and read (bonnie++)

Ran dmsetup to round robin the devices then failed to mount the volume as shown above.

 

 

 

Brian Urrutia

System Administrator

Price Communications Inc.

 


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]