problem restarting cluster - please help

[Fabio Coatti]

I need some help with a problem that I have with mysql cluster.
If I shut down cleanly the cluster, say using "shutdown" command from ndb-mgm, I'm unable to restart cluster: I've to launch ndbd with --initial option and recover from backup.
I've 2 storage nodes, 2 api nodes (mysql) active but several configured (for restoring backups) and one mgm node.
I've one storage and one api nodes on the same machine (two dual opterons) and one mgm-node, (i386).
All mysql installed are 4.1.12

If I shut down one storage node, I can restart it and the cluster is working in a few seconds. If I shot down both storage nodes, the restart sequence stops at phase 4, the logs is here:

2005-06-27 23:31:26 [MgmSrvr] INFO -- Node 3: DICT: index 111 activated done
2005-06-27 23:31:26 [MgmSrvr] INFO -- Node 3: DICT: index 112 activated done
2005-06-27 23:31:26 [MgmSrvr] INFO -- Node 3: DICT: index 113 activated done
2005-06-27 23:31:26 [MgmSrvr] INFO -- Node 3: DICT: index 114 activated done
2005-06-27 23:31:26 [MgmSrvr] INFO -- Node 3: DICT: index 115 activated done
2005-06-27 23:31:26 [MgmSrvr] INFO -- Node 3: DICT: index 116 activated done
2005-06-27 23:31:42 [MgmSrvr] ALERT -- Node 1: Node 3 Disconnected
2005-06-27 23:31:42 [MgmSrvr] INFO -- Mgmt server state: nodeid 3 freed, m_reserved_nodes 0000000000000016.


and the sequence restarts from first phase, no change if both nodes are restarted toghether.

The configuration seems ok, due to the fact that in normal operations all works fine and if I kill one storage node I can restart it without problems.

So my question are:
1) Is possible to restart a dual-node cluster after a clean shutdown without recovering from backup?
2) If this is possible, can someone give me an hint on what I'm doing wrong?

Many thanks in advance.