| Bug #42704 | "Slave_IO_Running" differs in test result (2) | ||
|---|---|---|---|
| Submitted: | 9 Feb 2009 16:27 | Modified: | 2 Jun 2009 17:05 |
| Reporter: | Joerg Bruehe | Email Updates: | |
| Status: | Closed | Impact on me: | |
| Category: | Tests: Replication | Severity: | S3 (Non-critical) |
| Version: | cluster 6.4.2 | OS: | Any |
| Assigned to: | Magnus Blåudd | CPU Architecture: | Any |
[20 May 2009 10:54]
Magnus Blåudd
Looking in the slave servers log file, I can see that the loop in include/wait_for_slave_io_to_stop.inc got "No such row" as result, when it then later compares that against 'No', the loop is terminated prematurely.
'No such row' can be returned from 'query_get_value', will double check that function to see why. If function looks ok, just change to SELECT 'Yes' = '<result from query_get_value on SHOW SLAVE STATUS>'
2 Query SHOW SLAVE STATUS^@
2 Query SELECT 'No' != 'No such row'
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2 Query SELECT '' = ''
2 Query SHOW SLAVE STATUS^@
2 Query SELECT NOT('No' = 'No') AND 3000 > 0
2 Query SHOW SLAVE STATUS
2 Query DROP TABLE t1
2 Query SHOW WARNINGS
[20 May 2009 11:01]
Magnus Blåudd
Hmm, someone has coded 'No such row' into the .inc file ...
[1 Jun 2009 15:08]
Magnus Blåudd
The slave IO thread reads the packet header from master and find it is a bigger packet than it can handle and will set an error code, so far it's fine and same as 6.3. For some reason "it" also call 'vio_was_interrupted', that function read errno which is 11 from an old operation -> yes, it was interrupted and thus a different path is taken where the slave tries to reconnect to the master instead of just erroring out as it should. While reconnecting it reset the Slave_IO_Running to YES for a brief time and thus this race condition occurs.
[1 Jun 2009 15:15]
Bugs System
Pushed into 5.1.34-ndb-7.0.7 (revid:magnus.blaudd@sun.com-20090601151346-9scmtzv2lzhueou4) (version source revid:magnus.blaudd@sun.com-20090601151346-9scmtzv2lzhueou4) (merge vers: 5.1.34-ndb-7.0.7) (pib:6)
[1 Jun 2009 15:16]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/75385
[1 Jun 2009 15:34]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/75389
[1 Jun 2009 15:35]
Bugs System
Pushed into 5.1.34-ndb-7.0.7 (revid:magnus.blaudd@sun.com-20090601153434-ai9qfmpwwmorscn0) (version source revid:magnus.blaudd@sun.com-20090601153434-ai9qfmpwwmorscn0) (merge vers: 5.1.34-ndb-7.0.7) (pib:6)
[2 Jun 2009 17:05]
Jon Stephens
Test failure only, no changes to document. Closed.
Description: Test fails on several platforms, especially Linux (RPM builds) and OS X, both x86 and x86_64 CPUs: ===== rpl.rpl_packet 'row' [ fail ] --- /PATH/mysql-test/suite/rpl/r/rpl_packet.result +++ /PATH/mysql-test/suite/rpl/r/rpl_packet.reject @@ -31,7 +31,7 @@ START SLAVE; CREATE TABLE `t1` (`f1` LONGTEXT) ENGINE=MyISAM; INSERT INTO `t1`(`f1`) VALUES ('aaaaa...aaaaa2048'); -Slave_IO_Running = No (expect No) +Slave_IO_Running = Yes (expect No) ==== clean up ==== DROP TABLE t1; DROP TABLE t1; mysqltest: Result length mismatch ===== How to repeat: Run the suite.