[DRBD-user] a few questions for the list

Jeff Buck jeffb at umci.com
Tue May 24 23:43:31 CEST 2005

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


I don't know if drbd will do the one -> many replication all on it's own
(it might), but I seem to remember reading a *long* time ago about
someone having done this using separate drbd devices, and using
mirroring on the "primary" server so that a write to drbd0 was also a
write to drbd1, drbd2, drbd3 etc.. (I think the example I saw was only
doing a 0 + 1 + 2 setup). I think they said that it was working for
them.

Under this setup though, the primary server would need the disk space
equivalent of #nodes * drbd volume size to be able to do all the
mirroring. I'm not sure, but you may be able to get around that by
running the drbd diskless, but primary for drbd1,2,3  on the primary
server. The downside would be that normal mirroring from local drive to
local drive is not nearly as smart as drbd is across the network, so any
error with the drbd connection is likely to force a full resync within
the "local" drive mirroring setup. Of course if you actually burnt
#nodes * volume size disk space setting it up, you'd not fail your
writes, and drbd could catch it's sync up in its own sweet time.

You might have to be a bit of a nutball to try the diskless extra
devices, but I felt I should mention the idea anyway. I'm not even sure
how, or if you could even do that. I've never tried using drbd in any
way other than its intended manner. 


On Mon, 2005-05-23 at 21:58 -0800, Allen Parker wrote:
> A pretty ascii picture of what I'd like to do:
> 
> drbd0:
> db   #==> web1
>      #==> web2
>      #==> web3
> drbd1
> web1 #==> web2
>      #==> web3
>      #==> db
> 
> My question:
> Is it possible with drbd, to do "live" one-to-many replication? If not,
> does anyone on the list know of any solutions that would allow a
> one-to-many replication, in other words a mirroring of data across all
> machines? Best case scenario, I'd be able to do many-to-many replication.
> 
> Thanks for your collective time,
> Allen Parker
> _______________________________________________
> drbd-user mailing list
> drbd-user at lists.linbit.com
> http://lists.linbit.com/mailman/listinfo/drbd-user




More information about the drbd-user mailing list