Bug #25913 rpl_ndb_dd_advance fails randomly
Submitted: 29 Jan 2007 12:40 Modified: 10 May 2007 7:47
Reporter: Michael Widenius Email Updates:
Status: Duplicate Impact on me:
None 
Category:MySQL Cluster: Cluster (NDB) storage engine Severity:S2 (Serious)
Version:5.1.16-beta OS:Linux (SuSE 11, x64)
Assigned to: Assigned Account CPU Architecture:Any
Tags: sr5_1

[29 Jan 2007 12:40] Michael Widenius 
Description:
rpl_ndb_dd_advance fails randomly with:

mysqltest: At line 515: query 'CALL tpcb.trans()' failed: 1205: Lock wait timeou
t exceeded; try restarting transaction

Have disabled test until this is fixed.

Additional notes:

rpl_ndb_dd_advance.log ends with:

***** Finsh the slave sync process *******
@the_epoch:=MAX(epoch)
<the_epoch>
@the_pos:=Position      @the_file:=SUBSTRING_INDEX(FILE, '/', -1)
<the_pos>       master-bin.000001
* 4. *
* 5. *
START SLAVE;
**** We should be ready to continue on *************

In the master.err file we have LOTS of messages like:
Waited WAITFOR_RESPONSE_TIMEOUT, continuing wait
(Someone should ensure we don't log so many of these...)

How to repeat:
compile-pentium64-debug-max
cd mysql-test
/mysql-test-run.pl --force --ps-protocol mysqld=--binlog-format=row
[25 Apr 2007 9:16] Guangbao Ni 
I just reproduce it in my local machine, but the result has a bit difference. one ndbd of master crashs in my test.
[25 Apr 2007 9:17] Guangbao Ni 
attach error and trace file, othter things are default.

Attachment: ndb_1_error.log (text/x-log), 569 bytes.
[25 Apr 2007 9:19] Guangbao Ni 
attach error and trace file, othter things are default.

Attachment: files.tar.gz (application/x-gzip, text), 64.24 KiB.
[29 Apr 2007 5:22] Guangbao Ni 
the ndbd crashes during  executing of tmp.p= (Page*)m_global_page_pool.getPtr(tmp.i); statement of Dbtup::get_dd_ptr().

According to the source code trace, i find that the member variable m_pgman ( req_struct->m_disk_page_ptr.i = m_pgman.m_ptr.i; in Dbtup::fireDetachedTriggers() ) of Dbtup class has the error value after alter a column with data into blob.But where should be the correct place to assign the right value? i still to trace it.
[30 Apr 2007 12:21] Tomas Ulin 
Fixed by -
Bug #28161 Detached triggers + DD and only MM update could cause node failure
[8 Aug 2007 16:39] Konstantin Osipov 
Tomas, the test case seems to be still disabled.