[Linux-cluster] complex gnbd setup example

[brianu]

Hello,

 

 

I am trying to setup a redundant system using gnbd ( linux-iscsi for now
fiber if I can figure this out).

 

Hardware and softwaredetails

HP MSA100 SAN

3 gnbd servers mounting the storage via linux-scsi in a backend lan &
exporting to 3 test nodes

All nodes import the gnbds, but in the event that one GNBD server dies I
want it to not affect the clients.

OS Centos 4 - kernel.org kernel 2.6.12

 

 

Problems-> 

1) multipath doesn't appear to work with GFS 6.1 obtained from CVS Stable
for latest kernel.org kernels.

2) LVM doesn't allow duplicate PVs ( the SAN to be created in the Volume
group)

i.e.

[root at dell-1650-31 ~]# vgcreate my_new_gfs /dev/gnbd/l108_sdata
/dev/gnbd/l109_sdata

  Found duplicate PV TbSaubq5Si2S1nrMbuMF620HPb6kFojE: using /dev/gnbd2 not
/dev/gnbd0

  Found duplicate PV TbSaubq5Si2S1nrMbuMF620HPb6kFojE: using /dev/gnbd2 not
/dev/gnbd0

  Volume group "my_new_gfs" successfully created

[root at dell-1650-31 ~]#

[root at dell-1650-31 ~]# vgchange -a y my_new_gfs

  Found duplicate PV TbSaubq5Si2S1nrMbuMF620HPb6kFojE: using /dev/gnbd2 not
/dev/gnbd0

  0 logical volume(s) in volume group "my_new_gfs" now active

[root at dell-1650-31 ~]#   

 

Ideas->

Heartbeat -linux-HA for a dynamic ip in which the nodes will mount  via
gnbd_import -I vip

 

Thoughts? Can someone suggest a better route, or how they might have
accomplished this ? the idea is a failover to another gnbd server should it
one go down.

I am open to hearing suggestions.

 

Thanks in advance.

 

 

Brian Urrutia

System Administrator

Price Communications Inc.

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20050817/b4110af3/attachment.htm>