Comments on: DRBD and Heartbeat for high availability on Linux

By: ilyas

ilyas — Sat, 17 Oct 2009 01:04:41 +0000

It’s great Lee! i’ve been trying to build a clustered server with two nodes. i have a problem, how whether the electricity comes down on primary node suddenly(e.g disaster or other).. is the synchronization happened..? when does it happen? how to solve it..?

thanks a lot!

By: Douglas Lochart

Douglas Lochart — Wed, 13 Feb 2008 16:09:52 +0000

First of thanks for the Tutorial it has been very helpful. If possible I would like some clarification. After you specify the ip addresses for eth1 (192.168.1.x) I see no mention of them in the config files save your mention of eth1 as the value for the bcast parameter. I see that this is the interface heartbeat uses to send the udp heartbeat packets. So the eth0 interface is used for DRBD syncing and eth1 is used for heartbeat. Since like your example I have a switched cable on my eth1 interface does it make sense to also add eth0 to the bcast line ?

bcast eth1 eth0

Or is that not necessary?

Also I am syncing a 1 terrabyte partition over a gigabit switch as of now. Should I tweak the syncer parameters up to 100M or more? Also the al-extents is set to 257 but the comment in the default config file says that this will handle a 1 gig active set. I see no definition of what that is. Since I ahve such a large partition to sync should I increase that ?

Thanks

Doug Lochart

By: Lee Hinman

Lee Hinman — Sat, 08 Dec 2007 19:12:24 +0000

I’ve updated the post with the correction, thanks Glynn!

By: Glynn Bird

Glynn Bird — Sat, 08 Dec 2007 16:21:55 +0000

A very useful tutorial. Many thanks. One small typo:

drbdisk::r0

should read

drbddisk::r0

on your line-by-line explanation.

By: Fernando Serer

Fernando Serer — Fri, 19 Oct 2007 07:19:25 +0000

Thanks Lee! I will try it this weekend

what i would like to be able to do is to have 2 servers, with apache in server A and mysql in server B.

They are connected to internet with eth0 and crossover linked with eth1.

Now apache is using the crossover link to read/write in mysql. But i’ve been reading that i could use this 2 servers to achieve high availability and each one would fail over the other service.

Is the active/active option in this link:

http://www.linux-ha.org/DRBD/GettingStarted

I’ve been googling on this option and I founded no more information than that page. All the information i found is about having a spare server waiting (like yours) and I would like to go one step beyond

I hope to be able doing it and i’ll tell you!

thanks again

By: Lee Hinman

Lee Hinman — Thu, 18 Oct 2007 20:21:04 +0000

Fernando: You should be able to let DRBD replicate through a crossover cable. I am fairly sure that it will be okay as long as both machines can see each other.

In the case of replicating over large distances, however, latency is going to be a bit problematic and you might not see the performance you are hoping for (depending on disk read/write numbers). A direct-connect crossover is going to be faster compared to a switch because of the lack of switch overhead.

Give it a shot and let me know how it work!

By: Fernando Serer

Fernando Serer — Thu, 18 Oct 2007 18:05:17 +0000

Good article and thank you for sharing your experience!

I’m interested in setting up something similar, but the crossover cable is connected to Gb NIC’s and the other local network NIC’s are connected to a 100Mbps switch.

Do you know if it’s possible (or if it’s safe) let drdb replicate through the crossover cable instead of the 100 Mbs nics?

In your example, the configuration of /etc/drbd.conf would be:

for lava2042
address 192.168.1.1:7788;

and for lava2138
address 192.168.1.2:7788;

I don’t know if this would be possible or safe to do. Thanks!

By: ragini

ragini — Tue, 10 Jul 2007 09:25:22 +0000

i want to setup two drbd resources on one machin having one primary server and two secondary server and both secondary should replicate same data.
can you help