[DRBD-user] Performance problems after upgrade to 0.7.8

[Felix Ide]

Hello list,

I've sent this message about a week ago, but at that time all of you
were very busy fixing the 0.7.9... bug ;) So I'm gonna try once again :)
Is this a 2.4 kernel problem, could it be better with 2.6 series?

Original post here:
After an upgrade from drbd 0.6.11+cvs to 0.7.8 we encountered serious
performance problems. Both nodes are identical, with 2x AMD Athlon 2200
MP, 2 GB RAM, 2x 3ware 7500 IDE Raid controller w/ 8x 120 GB HD,
connected directly via Gigabit Eth (Intel PRO/1000, MTU 9000).
SuSE 8.2, kernel 2.4.21-144-smp

After the update (internal metadata) the first, forced sync was only
about 4 - 5 MB/s in contrast to about 50 MB/s with 0.6.11+cvs! Even read
performance seems to suffer, but this was only a subjective impression. 
After the sync (> 24 hours!) write performance had also the same bad
values.

The only solution was to downgrade to 0.6 again because employees
complained about the slow server :(

What shall I do to track down the problem?

Thanks a lot,
Felix

Configuration was:
resource drbd0 {
    protocol               C;
    on linux1 {
        device           /dev/drbd0;
        disk             /dev/sda5;
        address          10.20.30.1:7788;
        meta-disk        internal;
    }
    on linux2 {
        device           /dev/drbd0;
        disk             /dev/sda5;
        address          10.20.30.2:7788;
        meta-disk        internal;
    }
    net {
        sndbuf-size      256k;
        timeout           60;
        connect-int       18;
        ping-int          18;
        max-buffers      128;
        on-disconnect    reconnect;
        ko-count           4;
        max-epoch-size   128;
    }
    disk {
        on-io-error      panic;
    }
    syncer {
        rate             100M;
        group              1;
    }
    startup {
        wfc-timeout        0;
        degr-wfc-timeout 120;
    }
}