Hi, I have a delicate question concerning pingd and attrd logic. I noticed that when connectivity stops/starts on both nodes with small delay, as happens in reality, there is a small delay in evaluation of the ping score which causes in part an unnecessary switch over.
More precisely here is the scenario: (a simplified version) Heartbeat 2.1.2 Two nodes CRM v2 ha.cf spawns pingd with default dampen=5, deadping>10 and keepalive=1 An active-passive resource R that may run on exactly one node. This resource R depends on the pingd generated score. No stickiness is used. 1. The resource R is initially running on node1. 2. Both nodes have ping connectivity. 3. Ping is stopped on node2 at time t1 4. Ping is stopped on node1 soon after, e.g. at time t2=t1+3 5. pingd on node2 reports it as dead around t1+deadping 6. pingd on node1 reports it as dead around t2+deadping 7. attrd requests update around ~t1+deadping+dampen 8. cibmon changes node1 value 9. The resource R is moved node1->node2 10. cibmon changes node2 value soon after 8. What can I do to prevent this? Is it a known issue? Am I doing something wrong? Your help is appreciated, Oren _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
