[Pacemaker] Cluster Amnesia problem in Pacemaker

chajo shakes1714 at yahoo.com
Fri Jul 30 12:15:06 UTC 2010


Hi, 
      i am a beginner in pacemaker and in the process of understanding it 
better. its usage on openais stack. while reading about cluster properties i 
came across this problem "clsuter amnesia". i am not sure this problem exists 
in Pacemaker ( i dont have setup right now to experiment this :( )

i am aware that when a node in a Pacemaker(on openais) has its configuration 
updated locally it is synchronized on all nodes of that cluster. and read in 
Pacemaker documentation that whenever a node joins cluster asks the node with 
the highest (admin_epoch, epoch,num_updates) tuple to replace the 
configuration on all the nodes. i understood this is the case when nodes are 
online.

cluster amnesia  is basically on a two node cluster (node A, node B), if node 
A is down and node B is up.. and on node B we add 2 new resources (make 
someother config change) and save the configuration. next we bring node B 
down. now when we start node A it comes up with old configuration ( it missed 
newly added 2 resources on node B). 

i would like to understand whether this problem exists with  with pacemaker? 

or some solution already exists in Pacemaker to take care of this problem?

i tried to search the forum with keyword "amnesia" couldn't get any results

am going to experiment this once i get setup in two days.


thanks in advance 
chajo





More information about the Pacemaker mailing list