Bug #51693 Starting a node during a partitioned start causes node failure

• Submitted: 3 Mar 2010 14:34
• Modified: 24 Mar 2010 20:57
• Reporter: Andrew Hutchings
• Status: Verified
• Category: MySQL Cluster: Cluster (NDB) storage engine
• Severity: S3 (Non-critical)
• Version: mysql-5.1-telco-6.3
• OS: Any
• CPU Architecture: Any

[3 Mar 2010 14:34] Andrew Hutchings

Description:
Error message observed:

Time: Tuesday 26 January 2010 - 12:46:47
Status: Temporary error, restart node
Message: System error, node killed during node restart by other node (Internal
error, programming error or missing error message, please report a bug)
Error: 2303
Error data: Node 4 killed this node because it replied StartFragRef error code: 0.
Error object: NDBCNTR (Line: 257) 0x0000000a
Program: ndbd
Pid: 18920
Trace: /local/cudb/mysql/ndbd/data/ndb_4_trace.log.4
Version: mysql-5.1.37 ndb-6.3.27a-GA

How to repeat:
.

[24 Mar 2010 20:57] Andrew Hutchings

How to reproduce:

1. Start 2 data node cluster
2. Insert some data (I did about 100MB)
3. Stop data node 1
4. Drop the table and re-insert the data
5. Stop data node 2
6. Start data node 1 with --nowait-nodes=2 (the other data node)
7. When data node 1 has completed phase 4 start data node 2

This has to be a non-debug build.  Debug builds seem to hit various ndbasserts too easily when trying this test case.