to reconnect:

  • on primary: drbdadm connect all

  • on secondary: drbdadm — --discard-my-data connect all

use cat /proc/drbd to see which are not connected, may have to connect only some, eg:

drbdadm connect r0##  or  ##drbdadm -- --discard-my-data connect all

to manually recover from split brain (split-brain):

select one node whose modifications will be lost:

  # resource is something like "r0" - perhaps "all" would work? haven't tried...
  drbdadm secondary <resource>
  drbdadm connect --discard-my-data <resource>

changes on the other node will be propogated to the node above:

for each resource that is StandAlone:

drbdadm connect <resource>

wait for synchronization…

to switch primary nodes:

on the secondary:

/etc/init.d/heartbeat stop

to check status:

drbd:

cat /proc/drbd

heartbeat:

crm_mon -1     # single-shot
crm_mon -i 1   # like top; refresh every second

updated: 2011-08-18