[Ocfs2-users] OCFS2 Multipath Configuration

Charlie Sharkey charlie.sharkey at bustech.com
Fri Mar 19 04:52:06 PDT 2010


When I mount an ocfs2 partition with the device mapper the mount command
shows a
/dev/dm-x device rather than a /dev/mapper device, but, I mount by
label. Ex:

   1. partition the drive  --  fdisk /dev/dm-0  (or can use /dev/sdx)
   2. reboot -- now have dm-0 and dm-1
   3. format the partition  --  mkfs.ocfs2  /dev/dm-1  -L myLabel  ....
/dev/dm-1
   4. mount by label. fstab:  LABEL=myLabel   /somedir  ocfs2  _netdev
0 0
   5. reboot 

The mount command will show /dev/dm-1 as the mount point. This is the
device
shown by 'ls -al /dev/disk/by-label' to be associated to myLabel.

This seems to work ok, if I pull a fibre connection the multipath
command will show the failed path and communication will continue
on another path. This could just be a Sles10 thing also ...



-----Original Message-----
From: ocfs2-users-bounces at oss.oracle.com
[mailto:ocfs2-users-bounces at oss.oracle.com] On Behalf Of Elliott Perrin
Sent: Thursday, March 18, 2010 9:58 PM
To: ocfs2-users at oss.oracle.com
Subject: Re: [Ocfs2-users] OCFS2 Multipath Configuration

I have seen node fences occur on all of our OCFS2 systems when the
O2CB_HEARTBEAT_THRESHOLD is set less than 31 on both FC and iSCSI
connected systems. In some testing I have done I found that all nodes
would panic with O2CB_HEARTBEAT_THRESHOLD set to any value less than 15
and we would see single node fencing with values between 15 and 30. With
the O2CB_HEARTBEAT_THRESHOLD set to 31 or greater we do not see a fence
occur when fabric disruption occurs, again on both iSCSI and FC. 

This is with a multipath configuration that includes 

polling_interval "5"
no_path_retry "5" 
features "1 queue_if_no_path"
failback "immediate"
selector "round-robin 0"

as some of the configuration variables used in the setup. 

As long as the output of mount (not mounted.ocfs2) shows that your file
system mount is from a /dev/mapper created multipath device then OCFS2
is using the multipath provided device. 


> -----Original Message-----
> From: ocfs2-users-bounces at oss.oracle.com [mailto:ocfs2-users-
> bounces at oss.oracle.com] On Behalf Of Sunil Mushran
> Sent: Thursday, March 18, 2010 2:11 PM
> To: David Johle
> Cc: ocfs2-users at oss.oracle.com
> Subject: Re: [Ocfs2-users] OCFS2 Multipath Configuration
> 
> Yeah.. mounted is a bit dumb. In the next release, it will recognize
> /dev/mapper devices. We still need to teach it to handle multipathing
> fully.
> 
> David Johle wrote:
> > I'm not sure about why mounted.ocfs2 is showing both the dm and the
> > sd devices for the same volume.  But this could all be very similar
> > to a problem I've experienced with OCFS2 & mount finding the right
> > device with multipathing.
> >
> > See the following thread for some more insight:
> > http://oss.oracle.com/pipermail/ocfs2-users/2009-March/003391.html
> >
> >
> >
> > At 04:12 PM 3/15/2010, you wrote:
> >
> >> Date: Wed, 16 Dec 2009 14:20:45 -0500
> >> From: "Nyburg, Daryl" <Daryl.Nyburg at utoledo.edu>
> >> Subject: [Ocfs2-users] OCFS2 Multipath Configuration
> >> To: <ocfs2-users at oss.oracle.com>
> >>
> >> Hello All,
> >>     I am having some problems configuring OCFS2 to use only the
> >> multipath device name.  We have been doing failover testing with
our
> >> ISCI SAN and as soon as we unplug one NIC the following messages
> appear
> >> in /var/log/messages and the system reboots.
> >>
> >>
> >> Dec 16 12:39:03 mcprac01 kernel: (56,6):o2hb_write_timeout:172
> ERROR:
> >> Heartbeat write timeout to device dm-34 after 120000 milliseconds
> >> Dec 16 12:39:03 mcprac01 kernel: (56,6):o2hb_stop_all_regions:1967
> >> ERROR: stopping heartbeat on all active regions.
> >> Dec 16 12:39:03 mcprac01 kernel: ocfs2 is very sorry to be fencing
> this
> >> system by restarting
> >>
> >>
> >> RHEL 5.3 Kernel 2.6.18-128.el5
> >> OCFS2 Versions
> >>   ocfs2console-1.4.3-1.el5
> >>   ocfs2-2.6.18-128.el5-1.4.2-1.el5
> >>   ocfs2-tools-1.4.3-1.el5
> >>
> >> Why does this show all device paths? Is there anyway to tell OCFS2
> to
> >> ignore the /dev/sd* devices ?
> >> $ mounted.ocfs2 -d
> >> Device                FS     UUID
> Label
> >> /dev/sdf1             ocfs2  2eaddbd4-fac6-4c83-a86d-357215730b23
> >> /dev/dm-25          ocfs2  2eaddbd4-fac6-4c83-a86d-357215730b23
> >> /dev/sdaj1           ocfs2  2eaddbd4-fac6-4c83-a86d-357215730b23
> >> /dev/sdd1            ocfs2  199f76a6-280a-46c6-812e-50712170a823
> >> /dev/dm-32          ocfs2  199f76a6-280a-46c6-812e-50712170a823
> >> /dev/sdai1           ocfs2  199f76a6-280a-46c6-812e-50712170a823
> >>
> >
> > _______________________________________________
> > Ocfs2-users mailing list
> > Ocfs2-users at oss.oracle.com
> > http://oss.oracle.com/mailman/listinfo/ocfs2-users
> >
> 
> 
> _______________________________________________
> Ocfs2-users mailing list
> Ocfs2-users at oss.oracle.com
> http://oss.oracle.com/mailman/listinfo/ocfs2-users

_______________________________________________
Ocfs2-users mailing list
Ocfs2-users at oss.oracle.com
http://oss.oracle.com/mailman/listinfo/ocfs2-users



More information about the Ocfs2-users mailing list