Bug 171622

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-GB; rv:1.7.12) Gecko/20050922 Fedora/1.0.7-1.1.fc4 Firefox/1.0.7

Description of problem:
The DLM doesn't clear its nodeid when it gets shutdown. 

If the cluster is shutdown and restarted - there is a strong possibility that nodes will have different nodeids than before. In this case the cluster will be running but all attempts to use the DLM will result in one of the following messages:

dlm: Can't bind to port 21064
dlm: cannot start lowcomms -98
or
dlm: cannot initialise comms layer
dlm: cannot start lowcomms -107

Although this sounds quite nasty I doubt it will happen very often. It requires that all nodes in a cluster be removed from that cluster but *not* shutdown and the DLM module *not* removed.

Version-Release number of selected component (if applicable):


How reproducible:
Always

Steps to Reproduce:
1. Join nodes to a cluster
2. cman_tool leave on all nodes
3. "cman_tool join" in a different order than before or change nodeIds using
   "cman_tool join -N<n>"
4. Attempt to use the DLM (eg start clvmd)

Of course this only happens if static nodeids are not specified in CCS.
  

Actual Results:  If the new cluster has the same nodeids as the old cluster you will see 

dlm: Can't bind to port 21064
dlm: cannot start lowcomms -98

If the node's old ID is not part of the new cluster then you will see:

dlm: cannot initialise comms layer
dlm: cannot start lowcomms -107


Expected Results:  flawless operation :)


Additional info:

A fix for this is in the STABLE branch of CVS.

The workaround is to always remove the DLM module when removing a node from the cluster if it is not to be shut down.