(ITS#6203) Problem & solution for back-ndb issue connecting to Cluster data nodes

[andrew.morgan@sun.com]

Full_Name: Andrew Morgan
Version: 2.4.16
OS: Ubuntu
URL: 
Submission from: (NULL) (90.193.196.76)


I am running MySQL Cluster with data nodes all running on the same host.

I configured, built and installed OpenLDAP with back-ndb but when I run "slapd
-d 5", it failed with "ndb_db_open: ni_cluster[0]->wait failed (1)".

Looking at the code in "back-ndb/init.cpp" this implies that it was able to
connect to the Cluster management node but only the first of the data nodes. The
cause of this is line 133:  "ni->ni_cluster[i]->wait_until_ready( 30, 0 );" -
the 30,0 means wait up to 30 seconds for first data node but then no time at all
for the others; the fact that the method returned 1 means that the first node
was found to be live. The second parameter should be set to a non-zero value to
give the other data nodes chance to show as being live. I changed the 0 to 30,
recompiled/installed and slapd came up correctly.