[DRBD-user] drbd 0.8 active/active and lvm - What stupid thing am I doing?

Lars Ellenberg Lars.Ellenberg at linbit.com
Thu Jan 18 12:39:35 CET 2007

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


/ 2007-01-17 00:17:28 +0000
\ Graham Wood:
> I've got a problem with trying to setup a cluster arrangement (using clvmd, from redhat 
> cluster) using v0.8rc1 on debian etch.  I'm hoping this is down to me being stupid - since 
> it'd be such a nice setup if it worked *grin*
> 
> Basically it doesn't seem that writes to node1 are copied to node2 properly, but writes to 
> node2 /are/ copied to node1.
> 
> Both machines are K7 CPUs with 1GB memory, and a pair of 80GB IDE disks mirrored.  /dev/md4 
> is then mirrored between the nodes using drbd to produce /dev/drbd0.
> 
> drbd runs over a separate VLAN to most other traffic - over bond1 (which is a pair of 8169 
> gigabit cards) - and the redhat cluster suite uses this as a secondary link (bond0 is the 
> primary link for the cluster as well as the public interface, since it's a very limited vlan 
> anyway). 
> 
> I don't think the HW is relevant (I ran memtest86+ and badblocks on both systems for 24 
> hours each before this, so I'm happy the HW is OK) - but it's included in case that's where 
> I'm being stoopid.
> 
> I've done various steps, and run md5 on the drbd device / lv on top of it at each step:
> 
> 	dd if=/dev/zero of=/dev/drbd0 bs=1M count=100 on Node1
> 		Node1: 2f282b84e7e608d5852449ed940bfc51  -
> 		Node2: 2f282b84e7e608d5852449ed940bfc51  -
> 
> 	dd if=/dev/random of=/dev/drbd0 bs=1M count=100 on Node1
> 		Node1: b096e3d5bc24215d56207d1d6b9ea1d6  -
> 		Node2: b096e3d5bc24215d56207d1d6b9ea1d6  -
> 
> 	pvcreate & vgcreate on Node1
> 		Node1: 7229ff35a7beb80cf9aca91a01cfadfa  -
> 		Node2: 7229ff35a7beb80cf9aca91a01cfadfa  -


right.
so, what does your lvm think which device it is using as a pv?

> 	lvcreate /dev/shared/mysql 1G lv
> 		Node1: f935fef26a51ea329892571a4a7e0922  -
> 		Node2: f935fef26a51ea329892571a4a7e0922  -

does it happen to use the "lower level device"?
check your lvm filter settings.
check what pvscan -vvvvv says it is using.
check what "dmsetup table" reports it is using as mapping device for
your lv. there should be something like
shared-mysql: 0 some-number linear 147:0 384
                                   ^^^
if that is not 147:0, you are screwed,
and lvm writes bypass drbd.

if that looks ok,
we have to dig deeper...

would be a pretty bad bug, if it where true... and definetely a show
stopper, since our testing here suggests its all smooth, and
we are about to release 8.0rc2 as final next week.

-- 
: Lars Ellenberg                            Tel +43-1-8178292-0  :
: LINBIT Information Technologies GmbH      Fax +43-1-8178292-82 :
: Vivenotgasse 48, A-1120 Vienna/Europe    http://www.linbit.com :
__
please use the "List-Reply" function of your email client.



More information about the drbd-user mailing list