Description of problem: If a node doesn't have a cluster.conf file, trying to connect to ccsd with a specific cluster name fails to retrieve the cluster.conf from another running node. If the connect is tried again, the file is retrieved. Version-Release number of selected component (if applicable): current cluster-project cvs How reproducible: Always Steps to Reproduce: 1. Create valid /etc/cluster/cluster.conf on one node 2. Start ccsd on that node 3. Start ccsd on a new node that has /etc/cluster but not /etc/cluster/cluster.conf 4. run 'ccs_test connect force <clustername>' on second node 5. rerun step 4 Alternately, replace 'ccs_test connect force <clustername>' with 'cman_tool join -c <clustername>' after modprobing the cman module. Actual Results: After step 4, I get this: Force is set. Setting cluster name to clu ccs_connect failed: No data available After step 5, I get this: Force is set. Setting cluster name to clu Connect successful. Connection descriptor = 0 Expected Results: The results of step 5 in step 4 Additional info: I ran ethereal on the first node, and the request is definitely getting to that node. You can filter on a udp packet on dest port 50007. For some reason ccsd just doesn't see it the first time.
With the recent changes to add multicast and IPv6 support, I haven't seen this. Please confirm.
I no longer see it. Seems to be fixed.
Updates with the proper version and component name.
Updates with the proper version and component name. Again, just love out tools.
Appears to be fixed.