[DRBD-user] iscsi + md0 = tell me why this is a bad idea

Artur (eBoundHost) artur at eboundhost.com
Wed Oct 22 00:23:51 CEST 2008

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

Sorry for top posting but I'm on the blackberry.

Another reason my proposal is no good is because of the cpu load on the client system. 

But I disagree with what you're saying. The resync would be definitely longer but mdadm is so easy to work with and does such a good job that for smaller deployments it may make more sense than drdb because it reduces the complexity of the system. 
------Original Message------
From: Greg Freemyer
To: Artur (eBoundHost)
Cc: drbd-user at lists.linbit.com
Subject: Re: [DRBD-user] iscsi + md0 = tell me why this is a bad idea
Sent: Oct 21, 2008 16:36

On Tue, Oct 21, 2008 at 4:41 PM, Artur (eBoundHost)
<artur at eboundhost.com> wrote:
> We're getting ready to test a drbd + nfs system in production and i started
> looking into iscsi.  Thought of something that i couldn't find an answer to
> on google, maybe i'm not looking hard enough (not on the first 5 results)
> Why not have 2 storage systems exporting iscsi and simply software md raid
> the exports on a target system?  This would take out complexity of drbd and
> heartbeat entirely, no?
> Only thing i can think against this setup is it pushes lots of bandwidth to
> both storage servers, but assuming that network is already in place this
> should not be a barrier.  And yes, it would require an additional 1gb nic on
> the target server (1 from each storage system)
> I'm definitely not going with this setup because drbd works great, but still
> looking for anything you can say about why my scenario would not be a good
> idea.
> --

I don't think there is anything fundamentally wrong with your solution, but...

What happens when the "target" system fails.  If you want
high-availability, you need to implement a failover cluster for it.
So now you have 2 servers on the frontend and 2 storage servers on the
backend.  I have an idea, lets combine the front-end and back-end
servers!!!  We could use this really cool technology called drbd!!!

FYI: I suspect the other big issue is what happens in your model in
the presence of a network failure, even if it is relatively short.  I
believe a typical RAID setup would immediately fail the iSCSI drive,
and then when it came back online it would require a very long slow
rebuild.  Drbd is much more graceful in that it maintains a bitmap of
out of date blocks and when the link comes back up, only those blocks
are sync'ed.

Greg Freemyer
Litigation Triage Solutions Specialist
First 99 Days Litigation White Paper -

The Norcross Group
The Intersection of Evidence & Technology

Best Regards,


More information about the drbd-user mailing list