[prev in list] [next in list] [prev in thread] [next in thread] 

List:       evms-devel
Subject:    Re: [Evms-devel] Re: RFP: EVMS
From:       Kevin Corry <corryk () us ! ibm ! com>
Date:       2002-01-02 21:49:32
[Download RAW message or body]

On Wednesday 02 January 2002 14:21, Matt Zimmerman wrote:
> On Wed, Jan 02, 2002 at 09:16:28AM -0600, Kevin Corry wrote:
> > Ah. Well, for starters, since those two partitions are in a raid1 device,
> > you will need to enable the MD plugin in EVMS in the kernel in order for
> > that to be discovered, and subsequently discovered by the LVM plugin. The
> > 0.2.4 release did not include Raid1 support, but I think it has since
> > been added to CVS. If you are feeling brave you can try it out. Keep in
> > mind, it is still *very* new, so you may just opt to not use that MD
> > device under EVMS until we have tested the MD plugins a bit more.
>
> Ah, so the volumes would only be re-exported through evms if it could
> recognize them all the way down through the meta device?  That is, finding
> an LVM volume group with an opaque device in it isn't sufficient?

Right. In EVMS, all volume discovery is done bottom-up from the disks. In LVM 
in particular, it will only be able to create VGs if it finds at least one of 
the PVs for that VG. If it creates a VG, and later figures out that a PV is 
missing (as in your case), about all it can do is complain that it wasn't 
found. The metadata for LVM doesn't allow the VG to know anything about that 
missing PV except its UUID. 

On the other hand, we are working on porting the AIX LVM to EVMS. The AIX 
metadata is such that every PV in a group has a copy of the metadata for 
every other PV in that group. So in the case where one PV is missing from a 
group, you would have a lot more information available about how the group 
should be assembled. Of course, the data on that missing PV is still gone, so 
this isn't a huge consolation. :-)

> > So there is definitely a discrepency between kernel and user-space
> > discovery for the segment manager. We will try to figure that out this
> > week.
>
> Let me know if there is any additional information I can gather to help you
> diagnose the problem.

Mark made some changes to the messaging in the kernel segment manager that 
might give us some more info on what is going wrong during discovery on your 
system. I would send you a patch, but enough stuff has changed since the 
0.2.4 release that I would suggest you grab all of the latest kernel code 
from CVS. Anonymous access instructions are on our web site 
(http://www.sf.net/projects/evms). Grab the "runtime" module, and copy the 
runtime/linux-2.4/drivers/evms and runtime/linux-2.4/include/linux/evms 
directories to your kernel source tree.

Try booting with those changes, and let me know what kernel messages you get. 
You may have to run the evms-rediscover utility as you did the last time.

Also, Mark said it might be helpful to get a copy of the MBR from your disk. 
Run this:
	dd if=/dev/scsi/host1/bus0/target5/lun0/disc of=/tmp/sdf.mbr bs=1k count=1
and mail me the sdf.mbr file. Hopefully we can get this figured out soon.

-Kevin

_______________________________________________
Evms-devel mailing list
Evms-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/evms-devel
[prev in list] [next in list] [prev in thread] [next in thread]