MySQL Bugs: #40961: for unconnected mysqld, it hangs with 'waiting for cluster startup signal'

Bug #40961	for unconnected mysqld, it hangs with 'waiting for cluster startup signal'
Submitted:	24 Nov 2008 0:23	Modified:	13 Apr 2012 9:52
Reporter:	Ionut Dumitru	Email Updates:
Status:	Verified	Impact on me:	None
Category:	MySQL Cluster: Cluster (NDB) storage engine	Severity:	S4 (Feature request)
Version:	mysql-5.1-telco-6.3	OS:	Linux (centos 5.2 (32bit))
Assigned to:		CPU Architecture:	Any
Tags:	from cluster 6.3.17, MySQL Cluster, ndb config, unconnected mysqld

Description:
this is not really a bug, but perhaps people may run into the same stupid situation i've been. i had a 9 ndb node cluster with 3 replicas and due to the higher need of ram, decided to do a 10 node 2 replicas cluster. so i changed config.ini to add the new node but for some weired reason, i didn't increase the api nodes (the new server needed to connect as well). i have the ndb_mgmd on a separate machine and i was doing my manual sql work from there as well. the problem that occured was that once i did the first restart of mysqld on the management node, it lost it's connection to the cluster and it's place was most likely grabbed by the new server i had introduced in the system. i ended up being unable to perform any operations on the cluster from my usual location. i lost a lot of time trying to figure out the issue.the only thing that looked suspicious was when i ran a 'show full processlist' and saw a master query something like 'waiting for cluster start signal'. it took me a lot of time to link this query to the mistake i did, due to several other problems i had in mind while creating the cluster configuration. doesn't really matter, but the point is, if there was a more clear message somehow it would have been more obvious. perhaps something like 'acquiring node id from mgm server'. the 'waiting fro cluster start signal' may be too ambiguous.

if you think this is worth pursuing, i'm glad i could point it out.

How to repeat:
try using a mysqld to access cluster data when it's left outside due to lack of api connections that can be made to the cluster (mgm config.ini has too few api entries)

Can you please attach logs, or (even better) add logs from ndb_error_reporter

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".

see bug#64929 for how to reproduce