the URL of a post in the thread I referred to in my previous post:
http://forums.mysql.com/read.php?25,75550,78814#msg-78814
(I apologize if the URL is not persistent)
An Update:
I've switched versions of each of the components a few times, and there may be other convoluting factors involved, but the ndbd process now lists (in the mgm console) as 'starting' in Phase 2 relatively soon after starting.
I left my setup alone for a while (~5hrs), and the ndbd process is still listed as 'starting' in Phase 2.
In case it helps:
Another bit of buggy behavior I've witnessed is that the mgmd does not always remove the data node from its list when it is asked to shut it down from the mgm console. When I killall from the data node, it always catches the termination. I don't have information on the relative frequencies, but I believe I've seen it terminate the data node and not reflect the change in the status listing (even after restarting the mgm console) and I've seen it claim to have terminated the data node, but neither reflect the termination in the cluster status listing or actually terminate the node.
Of course, all of that in the context that I've never actually had a data node fully 'started'!
Is there anything I can do to facilitate my would-be advisors? Would it help if I tried to capture 'strace' data and post that?