RE: [Xen-users] drbd and xen problem: disk not accessible

["Ross S. W. Walker" <rwalker@xxxxxxxxxxxxx>]

Ross S. W. Walker wrote:
> Marco Strullato wrote:
> > > > disk = [ "drbd:r0,xvda,w" ]
> > >
> > >             ^^^^^^^
> > >  Where did you get this config info from, as far as I
> > >  know it should be:
> > >
> > >  disk = [ "phy:drbd0,xvda,w" ]
> > 
> > 
> > I get this from this guide:
> > http://www.drbd.org/users-guide/s-xen-configure-domu.html
> > 
> > 
> > > Setting drbd resources per-lv will become very messy
> > >  over time as the number of guests increase. I advise
> > >  replicating PVs between hosts and adding the drbdX
> > >  resources to shared VGs. If you maintain all your
> > >  VMs in xenstore then the VM config should migrate
> > >  between machines so you don't need to have the VM
> > >  config present on both machines beforehand and
> > >  therefore you will not be able to start the same
> > >  VM on 2 machines at once and therefore shouldn't
> > >  need to run CLVM which would get in the way of VM
> > >  migration.
> > >
> > 
> > This is interesting: could you be more precise please?
> > How do you suggest me to set up the envirorment?
> 
> You will need to do this with Xen 3.2 as I do not
> believe that RH/CentOS libvirt will migrate the VM
> config!
> 
> Drbd setups are really only good for 2 servers, if you
> want more Xen servers then 2 then look into an iSCSI
> implementation.

You can by the way convert this existing setup to iSCSI
by turning these 2 servers into iSCSI targets and
exporting the /dev/drbd0 PV on each to say Xen blade
servers setup with multi-path fail-over (don't do it
round-robin as the drbd latency and io ordering will
get in the way!).

> So, say we have 2 servers, they have 2 SATA disks in
> a mirror for OS, and 4 SATA disks in a RAID5/6/50/60/10
> for guests. Both servers identical.
> 
> Lets say /dev/sda is OS mirror and /dev/sdb is guest
> RAID (could be /dev/mdX for OS and /dev/mdX+1 for RAID
> if doing software RAID).
> 
> Create a drbd resource out of /dev/sdb on both hosts
> set as primary-primary and sync it up fully. Test it
> make sure drbd is properly setup and performing as it
> should.
> 
> NOTE: You will need to keep the meta-data on a separate
> disk in drbd! I advise creating a LV on the OS mirror
> and putting the drbd meta-data there.

After some thought, this isn't a requirement, but
adviseable anyway to allow proper VG resizing and to
make sure the meta-data io doesn't collide with the
disk io.

> Next make /dev/drbd0 a pv with pvcreate, create a VG
> out of the PV, then on the other host do a pvscan, and
> import the VG there.
> 
> Then you can create an LV on host1, lvscan on host2
> and it should appear there. Test it out back and
> forth until your comfortable it works as advertised.
> 
> Create a domU config file and test it out with your
> first LV, make sure it boots and installs correctly,
> then when you are comfortable with it import it into
> the xenstore with 'xm new <config file>', after that
> the VM should appear in 'xm list' all the time until
> it is migrated to another host, whether it is up or
> down it will show up in the list.
> 
> Then bring it up on host 1 with 'xm start <dom name>'
> when the host is up and running you can do a
> 'xm migrate <dom name> <host name>' and it should
> just move over to the second host.
> 
> Once you are comfortable this works you might want to
> look into some kind of VM load-balancing/fail-over
> application. Maybe someone has a nice heartbeat with
> load balancing configs and scripts put together. The
> idea is to have the virtual machine servers self
> managing.
> 
> -Ross

______________________________________________________________________
This e-mail, and any attachments thereto, is intended only for use by
the addressee(s) named herein and may contain legally privileged
and/or confidential information. If you are not the intended recipient
of this e-mail, you are hereby notified that any dissemination,
distribution or copying of this e-mail, and any attachments thereto,
is strictly prohibited. If you have received this e-mail in error,
please immediately notify the sender and permanently delete the
original and any copy or printout thereof.


_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users