[DRBD-user] Unexpected behaviour (Consistent vs Inconsistent)

[Paolo Dina]

Good morning, I'm using DRBD 0.7.17 with Heartbeat. When both the nodes
are running, the system works fine: n1 is in its primary state, n2 in
secondary state and data are replicated. Playing with heartbeat I
verified that the same correct behaviour happens with n1 as secondary
and n2 as primary.

To gain confidence with the system I'm doing some simulation, trying to
predict the final result.

I copied some hundreds of mb on drbd device in primary state on n1, and
during the copy n2 has been powered off (gracefully one time, not
gracefully in another test). n1 detects the situation and /proc/drbd
says that devices are Primary/Unknown. So far so good.
Once the copy has finished I have shutdown n1 and then re-started n2. At
this point there are two different versions of data, with n1 (which is
down) owning the more recent version.

I'd expect that when n2 comes up, DRBD recognizes that the device has
outdated data and mark it as Inconsistent. Instead ld is Consistent.

I'd like to understand why this happen and if it possible to avoid a
such situation.

Thanks for any clarification,
Paolo Dina


This is /etc/drbd.conf:

global {
  minor-count 1;
  dialog-refresh 1;
}
resource drbd0 {
  protocol C;
  incon-degr-cmd "halt -pf";

  startup {
        degr-wfc-timeout 20;
        wfc-timeout 30;
        }

  disk { on-io-error detach; }

  syncer {
    rate 100M;
    group 1;
    al-extents 257;
  }
  on n1 {
    device     /dev/drbd0;
    disk       /dev/hdb;
    address    192.168.0.241:7788;
    meta-disk  /dev/hda2[0];
  }
  on n2 {
    device    /dev/drbd0;
    disk      /dev/hdb;
    address   192.168.0.242:7788;
    meta-disk /dev/hda2[0];
  }
}