drbd with N>2 nodes?

Philipp Reisner philipp at linuxfreak.com
Sun Apr 2 23:27:25 MDT 2000

Am Son, 02 Apr 2000 schrieb Lars Marowsky-Bree:
>Good morning everyone,
>is work being done to extent drbd above 2 nodes? 
>I could live with the fact if it initially only allowed one of the nodes to
>If we require a dedicated TCP/IP link from the master to each "slave", this
>may be quite simple(?) to implement. Of course, it would be way cooler if we
>used mcast and didn't retransmit the data N-1 times...


What happened so far:

 October 1999:  Started drbd.
 November:      drbd-0.1. clueless.
 Januar 2000:   At this time I was *really* understanding the issues.
                I wrote the paper and drbd-0.5
 Feb-April:     Debugging and writing my theses.

My current roadmap looks like this

 1) Port drbd to 2.4.x (2.3.99-x)
 2) Test drbd with LVM and software raid.
 *) Include GFS support
 *) Write more automatic testing scripts
 3) Release drbd 1.0
 *) Graduate from University.          (June?)
 4) Update the drbd-paper (I have learned a lot in the process of writing
    drbd so far)
 5) Sign up at some Linux company      (August?)
 6) Working out the way drbd 2.0 could work (based on UDP multicast,
    supporting more than two nodes) 
 7) Working on drbd 2.0

I think it's a good idea to do GFS support before starting with the UDP
multicast based protocol.


More information about the Linux-HA mailing list