Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
>>>>> On Wed, 05 Apr 2006 17:41:04 -0400, Ugo Bellavance <ugob at camo-route.com> said:
> 3- Start drbd
> # service drbd start
> 4- Make the primary primary
> # drbdadm -- --do-what-I-say primary all
In a guide, I would not use "all" but the resource name, in your
example drbd0. I'd probably choose an example that does not confuse
the reader: if I'd demonstrate setting up mysql, I'd call it
drbd_mysql.
> 5- Check the status
> # watch cat /proc/drbd
> 6- Once in sync, create a filesystem on the drbd device (on both servers)
> # mkfs -t ext3 /dev/drbd0
Only on primary is enough. Don't know if it breaks something to do it
on both.
> 7- Create a mount point (both servers)
> # mkdir /drbd
> 8- Mount the drbd device on it (only on the primary)
> # mount /dev/drbd0 /drbd
> Finally, integrate with Heartbeat (which I haven't understood yet either.
It means that you do not ever do steps 3-8 again yourself. Steps 3-8
are ok to learn the basics and play with them, but as soon as
heartbeat does it for you, you let it alone. So
9 - Let drbd alone because it shall be owned by heartbeat (on
current primary)
# umount /drbd; drbdadm down drbd0;
As heartbeat usually involves starting a service that uses the mounted
disk you would now disable that service from being started by the OS
10 - Disable OS from starting mysql (on both nodes)
# rm /etc/rc?.d/[SK]*mysql
Talk about the need to prevent your upgrade mechanism (here yum) from
reinstalling those start/stop symlinks.
11 - Configure heartbeat (on both nodes)
# involves authkeys, haresources, and ha.cf; see
# http://linux-ha.org/DRBD/GettingStarted
12 - Start heartbeat (on both nodes)
This will bring up everything in the desired order and only on the
machine that is allowed to do it: make the DRBD primary, mount the
disk, ifconfig an ethernet alias interface, start mysql.
13 - test failover (on current primary)
# hb_standby # http://linux-ha.org/hb_standby
# hb_takeover # http://linux-ha.org/hb_takeover
14 - test failover (on the other node)
# hb_takeover
# hb_standby
> Am I ok? One question I had is:
> I've seen some examples using partition devices instead of disk
> devices in drbd.conf. How do they do that? What kind of partition do
> they create on the disk to be able to put a drbd device over it? My
> example uses /dev/sdb, which is the 2nd SCSI disk. I didn't create
> any partition on it, in fact, the partition is the drbd device and I
> created the FS on the drbd device. Is that OK? Would it be better to
> use a partition instead? How?
pass
--
andreas