Bug #4586 System restart fails after graceful shutdown
Submitted: 17 Jul 2004 19:18 Modified: 10 Aug 2004 10:55
Reporter: Johan Andersson Email Updates:
Status: Closed Impact on me:
Category:MySQL Cluster: Cluster (NDB) storage engine Severity:S1 (Critical)
Version:mysql 4.1 bk src OS:Linux (RH9)
Assigned to: Jonas Oreland CPU Architecture:Any

[17 Jul 2004 19:18] Johan Andersson
Cluster was shutdown gracefully. ALL STOP from mgmtclient.
Cluster failed to restart:

Date/Time: Friday 16 July 2004 - 03:44:01
Type of error: error
Message: System error
Fault ID: 2303
Problem data: Unable to find restorable replica for table: 0 fragment: 1 gci: 4783
Object of reference: DBDIH (Line: 8577) 0x0000000a
ProgramName: NDB Kernel
ProcessID: 2603
TraceFile: NDB_TraceFile_1.trace

We could get around this with restarting with --initial on that node.
Is everything really flushed to disk when the nodes shutdown (gracefully)???

4 db nodes
DataMemory: 256MB
IndexMemory: 200MB 
Load the table full. The table was loaded from the mysql client and the mysql client got an err mess saying table full.

Performed a graceful shutdown followed by restart.
We got this a few times and the what you seen in the error log emerged on  different nodes.

How to repeat:
We got this quite often.
[5 Aug 2004 18:04] Jonas Oreland

I haven't reproduced the exact error that you stated.

But I fixed 3 bugs in SR so that testSystemRestart -n SR_FULLDB now
works flawlessly, for me :-)
[10 Aug 2004 10:55] Jonas Oreland
Note, if index gets full problem still exists, but data should work