[Linux-cluster] Issue with clvmd - Is it really bug??

[Christine Caulfield]

Theophanis Kontogiannis wrote:
> Hello,
> 
>  
> 
> I have a 2 node cluster at home with CentOS 5 running on 64bit AMDx2 
> with DRBD
> 
>  
> 
> 2.6.18-92.1.6.el5.centos.plus
> 
> drbd82-8.2.6-1.el5.centos
> 
> lvm2-2.02.32-4.el5
> 
> lvm2-cluster-2.02.32-4.el5
> 
> system-config-lvm-1.1.3-2.0.el5
> 
>  
> 
> I do not know if my problem is directly related to 
> http://kbase.redhat.com/faq/FAQ_51_10471.shtm and 
> https://bugzilla.redhat.com/show_bug.cgi?id=138396
> 
>  
> 
> I do:
> 
>  
> 
> pvcreate --metadatacopies 2 /dev/drbd0 /dev/drbd1
> 
> vgcreate -v vg0 -c y /dev/drbd0 /dev/drbd1
> 
> lvcreate -v -L 348G -n data0 vg0
> 
>  
> 
> Then I reboot.
> 
> The LV never becomes available.
> 
>  
> 
> If I try
> 
>  
> 
> vgchange -a y
> 
>  
> 
> I get
> 
>  
> 
>   Error locking on node tweety-1: Volume group for uuid not found: 
> 7Z9ra5zee3ZK7pNpfsblvtMOWXhgkZVEiJrzRQshaaiN5JKtJtkPDkQWfFXYKVVa
> 
>   0 logical volume(s) in volume group "vg0" now active
> 
>  
> 
> If I do
> 
>  
> 
> clvmd –R
> 
>  
> 
> Then with
> 
>  
> 
> vgchange –a y vg0.
> 
>  
> 
> the LV becomes available.
> 
>  
> 
> Is this really related to the above mentioned bug?
> 
>  
> 
> How can I make the LV become available during boot up without any 
> intervention?
> 
>  
> 
> Thank you all for your time,

As you're using drbd for the PV, I think it might be to do startup 
ordering. If drbd is started AFTER clvmd then it won't see the devices, 
and you'll get exactly the symptoms you describe.

if you can, move drbd to before clvmd, or clvmd after drbd. Or, failing 
that, put the extra commands you used above into their own startup script.
-- 

Chrissie