[DRBD-user] Preventing initial full sync on new install

Christophe Boyanique tof+drbd at raceme.org
Sun Nov 25 21:43:45 CET 2007

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

Lars Ellenberg a écrit :

> as you quoted above, and as Stefan Seifert pointed out,
> the device _is_ usable right away.

Thanks to Stefan and you for your answers.

> especially in your case, what I'd do is:
>  unplug that external bay from one of the nodes.
>  attach some new storage to that node.
>  configure drbd with external meta data to use the existing data on the
>  node that still sees it, make it primary there, and start your
>  services.
>  then start drbd on the node with the newly attached storage,
>  and let it sync up.
>  now, understand and configure heartbeat.

In fact I can't preserve data:

There is two servers connected to an external IBM bay (with 2 SCSI
cables) and the bay is configured with only one partition. That means
that both servers connects to the same bay partition (only the master
mounts it) which is not supported by the bay and leads to data
corruption if both servers access data on the same time (which occured
several times in the past).

The client asked advice to IBM and "discovered" that the architecture
was clumsy. This is why we will have to backup data, add new disks to
the bay, divide it into two partitions (one for each server) and then
restore data.

So each server will have a partition on the bay synced by DRBD.


More information about the drbd-user mailing list