[Linux-ha-dev] client sending messages stuck heartbeat
gshi at ncsa.uiuc.edu
Fri Oct 8 02:03:25 MDT 2004
I think I found a bug.
With CVS head, replace the attached sender.c with lib/hbclient/api_test.c, compile it.
1. start heartbeat in both machines (A and B)
2. start api_test in one machine (A)
api_test will keep sending ordered messages to the cluster one for each second,
therefore this is not flow control problem
after sometime (~2 minutes in my machines), B decared A dead and took over A's resources
there is nothing wrong in A's log.
3. kill api_test in A, A started to "wake up" and find itself dead, then both machine restarted.
It seems master process was stuck in somewhere related with the client.
here is ha.cf
bcast eth0 # Linux
bcast eth1 # Linux
apiauth ping gid=haclient uid=gshi,root
apiauth ccm gid=haclient uid=root
apiauth evms gid=haclient uid=root
apiauth ipfail gid=haclient uid=gshi,root
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
More information about the Linux-HA-Dev