| Bug #18969 | race condition involving slave and mysqltest. was rpl_insert_id test case fails | ||
|---|---|---|---|
| Submitted: | 10 Apr 2006 16:46 | Modified: | 3 Oct 2006 20:23 |
| Reporter: | Brian Pontz | Email Updates: | |
| Status: | Closed | Impact on me: | |
| Category: | MySQL Server: Tests | Severity: | S3 (Non-critical) |
| Version: | 5.1.9-pre | OS: | Linux (RHAS3 x86) |
| Assigned to: | Magnus Blåudd | CPU Architecture: | Any |
[10 Jul 2006 14:08]
Andrei Elkin
The failure is not about replication. To prove there is a slave.err e.g https://intranet.mysql.com/~knielsen/getlog.pl?dir=mysql-5.1-new-rpl&entry=mats@romeo.none... CURRENT_TEST: rpl_insert_id 060708 8:43:55 InnoDB: Started; log sequence number 0 184637 060708 8:43:55 [Note] /data/pushbuild/pb/mysql-5.1-new-rpl/mats@romeo.none-20060708062142/mysql-5.1-new-rpl-clone/sql/mysqld: ready for connections. Version: '5.1.12-beta-debug-log' socket: '/dev/shm/pbtmp-ps_row-6/slave.sock' port: 10062 Source distribution 060708 8:43:55 [Note] SCHEDULER: Manager thread booting 060708 8:43:55 [Note] SCHEDULER: Loaded 0 events 060708 8:43:55 [Note] SCHEDULER: Suspending operations 060708 8:43:55 [Note] Slave SQL thread initialized, starting replication in log 'FIRST' at position 0, relay log '/dev/shm/var-ps_row-6/log/slave-relay-bin.000001' position: 4 060708 8:43:55 [Note] Slave I/O thread: connected to master 'root@127.0.0.1:10060', replication started in log 'FIRST' at position 4 060708 8:43:55 InnoDB: Error: table test.t1 already exists in InnoDB internal This is quite probable that rpl_foreign_key_innodb.test has not cleaned table test.t1 for a race condition of slave mysqld and mysqltest. The problem is in that the included test extra/rpl_tests/rpl_foreign_key.test ends with connection master; SET FOREIGN_KEY_CHECKS=0; DROP TABLE IF EXISTS t1,t2,t3; SET FOREIGN_KEY_CHECKS=1; i.e without sync with the slave so that there is no guarantee slave mysql DROPed the tables. A method to avoid that already is widely used is to complete the test with sync_slave_with_master; or equivalents. TODO: We need to instuct developers on that separetly! Since this is not the only single test that has the flaw, the full list of vulnerable tests besides mentioned parent rpl_foreign_key_innodb.test includes - check me out! - rpl_auto_increment_11932.test rpl_misc_functions.test rpl_multi_delete2.test rpl_ndb_bank.test rpl_ndb_basic.test rpl_ndb_dd_basic.test rpl_ndb_idempotent.test rpl_ndb_load.test The failure is not about replication. To prove there is a slave.err e.g https://intranet.mysql.com/~knielsen/getlog.pl?dir=mysql-5.1-new-rpl&entry=mats@romeo.none... CURRENT_TEST: rpl_insert_id 060708 8:43:55 InnoDB: Started; log sequence number 0 184637 060708 8:43:55 [Note] /data/pushbuild/pb/mysql-5.1-new-rpl/mats@romeo.none-20060708062142/mysql-5.1-new-rpl-clone/sql/mysqld: ready for connections. Version: '5.1.12-beta-debug-log' socket: '/dev/shm/pbtmp-ps_row-6/slave.sock' port: 10062 Source distribution 060708 8:43:55 [Note] SCHEDULER: Manager thread booting 060708 8:43:55 [Note] SCHEDULER: Loaded 0 events 060708 8:43:55 [Note] SCHEDULER: Suspending operations 060708 8:43:55 [Note] Slave SQL thread initialized, starting replication in log 'FIRST' at position 0, relay log '/dev/shm/var-ps_row-6/log/slave-relay-bin.000001' position: 4 060708 8:43:55 [Note] Slave I/O thread: connected to master 'root@127.0.0.1:10060', replication started in log 'FIRST' at position 4 060708 8:43:55 InnoDB: Error: table test.t1 already exists in InnoDB internal This is quite probable that rpl_foreign_key_innodb.test has not cleaned table test.t1 for a race condition of slave mysqld and mysqltest. The problem is in that the included test extra/rpl_tests/rpl_foreign_key.test ends with connection master; SET FOREIGN_KEY_CHECKS=0; DROP TABLE IF EXISTS t1,t2,t3; SET FOREIGN_KEY_CHECKS=1; i.e without sync with the slave so that there is no guarantee slave mysql DROPed the tables. A method to avoid that already is widely used is to complete the test with sync_slave_with_master; Since this is not the only single test that has the flaw, the full list includes besides the parent rpl_foreign_key_innodb.test - check me out! - rpl_auto_increment_11932.test rpl_misc_functions.test rpl_multi_delete2.test rpl_ndb_bank.test rpl_ndb_basic.test rpl_ndb_dd_basic.test rpl_ndb_idempotent.test rpl_ndb_load.test rpl_ndb_multi.test rpl_rbr_to_sbr.test rpl_relayspace.test rpl_rewrt_db.test rpl_row_basic_11bugs.test rpl_row_inexist_tbl.test rpl_row_stop_middle.test rpl_row_stop_middle_update.test (make it uniform to DROP only on master - and user replication to drop for granted on slave - always at least one line of code less) rpl_row_trig003.test rpl_slave_status.test (same comment on uniform) rpl_temporary.test rpl_truncate_7ndb.test and called from top level tests: extra/rpl_tests/{ rpl_ddl.test, rpl_failed_optimize.test, rpl_flsh_tbls.test, rpl_loaddata.test, rpl_row_func003.test, rpl_row_sp007.test, rpl_row_UUID.test,rpl_stm_EE_err2.test, rpl_truncate_helper.test} Mats, I am asking to readdress the issue to QA since the issues carries rather intergral character to deal with many tests.
[26 Sep 2006 11:50]
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/12514 ChangeSet@1.2285, 2006-09-26 13:49:42+02:00, msvensson@neptunus.(none) +1 -0 Bug#18969 race condition involving slave and mysqltest. was rpl_insert_id test case fails - As the slaves are restarted for each testcase, take the opportunity to restore their databases to a known state with the help of the snapshot(s) - Count max number of slaves used in testcases - Use copy_install_db to speed up "Installing db" phase
[3 Oct 2006 20:02]
Chad MILLER
Available in 5.0.26.
[3 Oct 2006 20:10]
Chad MILLER
Available in 5.1.12-beta.
[3 Oct 2006 20:23]
Paul DuBois
Test suite change only. No changelog entry needed.
Description: rpl_insert_id [ fail ] Errors are (from PATH/mysqltest-time) : mysqltest: In included file "./extra/rpl_tests/rpl_insert_id.test": At line 47: could not sync with master ('select master_pos_wait('master-bin.000001', 1415)' returned NULL) (the last lines may be the most important ones) Result from queries before failure can be found in r/rpl_insert_id.log Killing Possible Leftover Processes How to repeat: run test case