On 16/08/2016 07:09, Craig Thompson wrote:
Basic DRBD link not syncing
First post to the list. I've tried using search engines to find this
info but haven't so far.
I have two servers. They each run Xen hypervisor to support guest
virtual machines. They each have multiple LVMs for the underlying
disk images of the virtual machines.
I have set up Machine A to be the Primary for Guest1. Machine B is
the Secondary for Guest1.
The config I have for each looks like this:
resource guest1 {
net { protocol C; }
device /dev/drbd3;
disk /dev/vg0/guest3;
meta-disk internal;
syncer { verify-alg "md5" ; }
on primary.host.domain {
address 192.168.1.10:7789;
}
on secondary.host.domain {
address 192.168.1.20:7789;
}
}
I can use drbdadm and bring the resources up. I can force a primary.
This initiates a full sync the first time.
However, if I emulate an outage on Machine B (secondary) for minutes
or hours and then bring it back online, the syncing doesn't continue
automatically.
I don't know what I'm missing in configuration(s). I may be missing
something in the overall picture. But what I WANT to have happen is
that DRBD sees when the link is down, pauses and then picks up syncing
after an outage on the remote end.
If I run a 'drbdadm verify guest1', then it will initiate a review,
sync the out of sync blocks and then say it's up to date.
But if I go to the Machine A and create a new file on guest1, then
shut down guest1 so that no other changes are made, and then I go to
Machine B and bring guest1 up, the newly created file is NOT there.
So, I'm missing some key component that KEEPS things syncing after the
initial sync is completed.
Help? Documentation pointers? FAQs? How-Tos?
Thanks in advance.
There was a similar question on the list just a few days ago, not sure
if that was you.
In any case, the best thing to check/verify is that your VM is accessing
/dev/drbd3 and that nothing (except DRBD itself) is accessing
/dev/vg0/guest3 (on both primary and secondary).
Any changes made to /dev/vg0/guest3 are unknown to DRBD, and so DRBD
will not know that they should be synced to the other node. It is the
same as creating a MD array with /dev/sda1 and /dev/sdb1 as /dev/md1 and
then writing to /dev/sda1 and complaining there is not consistent data
on /dev/sdb1.
I hope that helps, if not, please see if you can include a step by step
process, with all commands you entered, all output generated from each
step, which will show the problem.
Regards,
Adam
_______________________________________________
drbd-user mailing list
[email protected]
http://lists.linbit.com/mailman/listinfo/drbd-user
