MySQL Bugs: #19005: A data node couldn't start successfully (phase 4 )

Bug #19005	A data node couldn't start successfully (phase 4 )
Submitted:	11 Apr 2006 13:12	Modified:	11 May 2006 14:55
Reporter:	Serge Kozlov	Email Updates:
Status:	No Feedback	Impact on me:	None
Category:	MySQL Cluster: Cluster (NDB) storage engine	Severity:	S2 (Serious)
Version:	5.1.9 (beta)	OS:	Linux (FC4)
Assigned to:		CPU Architecture:	Any

Description:
For some cases data node can't started and it always stays in phase 4:
ndb_mgm> all status
Node 2: started (Version 5.1.9)
Node 3: starting (Phase 4) (Version 5.1.9)

How to repeat:
1. Start cluster with 2 ndbd/2 replicas.
2. Run './load_tpcb.pl ndb16 3306 root BLANK ndb' and wait while 'Loading accounts table' message will appears.
3. Stop non-master data node.
4. Stop the script (Ctrl-C) if one wasn't interrupted.
5. Run ./load_tpcb.pl again (repeat point 2).
6. Start data node (it was stopped in point 3) without --initial  option.
Now we've following case:
Master node has 'new' data because the script was restarted (db and all tables recreated).
Non-master node has 'old' data.

1) When  refering to transactions (select,insert/update/delete) 
     there is no such concept as "master", it's an active/active cluster.

     The "master" concept is only used for ddl, checkpoints and other cluster
     internal protocols.

2) How did you stop node?
3) Does load_tpcb.pl perform ddl?
    If so, then this is a duplicate bug.
4) please upload clusterlog

1. Probably i'm wrong but SHOW command always select one node as 'master'. I meant this term.
2. Stop via XXX STOP command.
3. What do you mean 'ddl' for load_tpcb.pl?

ddl = create/drop/alter 
Also, please upload cluster log when it fails

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".