Mailing List Archive: Interface failure due to DRBD (?)

Hi,

I am experiencing the following problem:

DRBD is configured and connected on node 1 and node 2.
Node 1 is formatted ext2 (over RAID1.)
Node 2 is formatted ext2 (non-RAID.)
Both partitions are the same size.
Node 1 has a Realtek 8139B.
Node 2 has a Realtek 8139B.
The local interfaces on both machines are configured correctly.

I run DRBD, with node 1 as primary and node 2 as secondary - and this is fine
so far. cat /proc/drbd on both machines shows correctly connected DRBD setups.

Mount /dev/nb0 on node 1 (primary)
Create a small test text file - unmount on primary, change node 1 to
secondary, and change node 2 to primary, and mount /dev/nb0 - I see the file,
it is intact -- All is good so far.

Back to node 1 as primary, I start copying files to it. This is when the real
trouble starts: After a few blocks, DRBD kills the local interfaces on each
machine (node 1 cannot ping node 2, etc) and it requires a simple ifconfig
ethX down, ifconfig ethX up, and it works again.

After DRBD renegotiates, it once again kills the interfaces.

I have, in the past, tested this exact same configuration, except, with a
Realtek 8029 in node 2. I successfully mirrored 3.5Gig of data.

Now, it could be a problem with the Realtek 8319B, but this card worked
before in node 1. Also, I have tried Intel EEPRO100's in both nodes, and the
same problem occured.

Any suggestions?
Thanks for your time,
Omar

--
Omar Kilani
Systems Administrator
Mail Call Couriers Pty Ltd

> Back to node 1 as primary, I start copying files to it. This is when the
real
> trouble starts: After a few blocks, DRBD kills the local interfaces on
each
> machine (node 1 cannot ping node 2, etc) and it requires a simple ifconfig
> ethX down, ifconfig ethX up, and it works again.
>
> After DRBD renegotiates, it once again kills the interfaces.
>
> I have, in the past, tested this exact same configuration, except, with a
> Realtek 8029 in node 2. I successfully mirrored 3.5Gig of data.
>
> Now, it could be a problem with the Realtek 8319B, but this card worked
> before in node 1. Also, I have tried Intel EEPRO100's in both nodes, and
the
> same problem occured.

I saw the same problem and reported it few weeks ago. I suspect a bug in the
network code somewhere. Perhaps going for 2.2.18 will help ??
I was using 2 Realtek too !!

My advice are :
- Change your internal ethernet card fgor 3Com, it worked for two of my
servers
- Never use ping on the private interface with the realtek (it kill its)
- Never reach 50 of charge (uptime report) it kill the connection ;*)

Thomas