Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
> I've got 4 drbd partitions on 4 SCSI disks (15000rpm): > - 1 for NFS/Apache > - 1 for Postgresql data > - 1 for Postgresql WAL > - 1 for Postgresql index > > I've got 3 x 1 GB Ethernet controllers on these servers. > > Can I use one Ethernet controller for all drbd syncs ? DRBD uses IP level stuff, therefore you can do anything you like at the network layer, and it'll just run over the top? > Or is it useful to split drbd syncs on 2 Ethernet controllers ? > (for example, sync drbd0 and drbd1 on eth1 and sync drbd2 and drbd3 on eth2). Why limit it to something "forced"? You want to get the maximum possible throughput, so just configure bonding at the kernel level, and let that do load balancing instead. I've got a cluster with a pair of gigabit cards doing the heartbeat. Instead of setting them up as 2 links, I've just created a single "bond1" that uses a dual active arrangement. In my case this is 2 crossover cables, but if you've got a more complicated network then this may not be as easy to carry out :) However, with switch support, you can do this under almost all circumstances. Graham