Bug #34528 rpl_ndb_basic fails since slave does not synchronize in time

• Submitted: 13 Feb 2008 16:00
• Modified: 27 Oct 2010 9:13
• Reporter: Sven Sandberg
• Status: Can't repeat
• Category: MySQL Cluster: Replication
• Severity: S2 (Serious)
• Version: mysql-5.1-telco-6.2
• OS: Any
• Assigned to: Jørgen Austvik
• CPU Architecture: Any
• Tags: cluster, deadlock, mysql-5.1-telco-6.2->, ndb, replication, rpl_ndb_basic, synchronize

[13 Feb 2008 16:00] Sven Sandberg

Description:
rpl_ndb_basic fails with a result contents mismatch when using a sleep to wait for the slave to reach deadlock.

The test produces a deadlock as follows:

  connection slave;
  modify table t1;
  connection master;
  modify table t1 in a conflicting way;

Then, it synchronizes the slave as follows:

  sleep 5;
  connection slave;
  show slave status;

However, in the pushbuild run linked above, 5 seconds was apparently not enough: the slave was still running when show slave status was issued. Hence the result contents mismatch.

How to repeat:
Read the logs from powermacg5 on
https://intranet.mysql.com/secure/pushbuild/showpush.pl?dir=mysql-5.1-new-rpl&order=472

Suggested fix:
Create a mysql-test/include/wait_for_slave_to_deadlock.inc (which should wait until the slave has a deadlock, similar to, e.g., wait_for_slave_sql_error.inc). Use that instead of 'sleep 5'.

[27 Oct 2010 9:13] Sven Sandberg

This bug refers to an old pb log that is not accessible any more. So it's not clear what the symptom is and there is not much we can do to fix it. Closing as "Can't repeat".

If there are new bugs in this test case, please open a new bug.