RSU processor remains hanging

[Seppo Jaakola]

Howto reproduce:

1. run one node cluster with RSU enabled
2. issue high rate SQL write load on the node, e.e.g by sqlgen
3. through another connection, send safe DDL, e.g CREATE INDEX for one of the affected tables

If bug is present, the DDL will remain hanging

This seems to be a regression from one of the fixes introduced for: lp856162 => create a regression test

[Seppo Jaakola (seppo-jaakola)]

Fix pushed in: http://bazaar.launchpad.net/~codership/codership-mysql/wsrep-5.5/revision/3769

[Seppo Jaakola (seppo-jaakola)]

Regression test failed for wsrep-5.6 branch. Following assert was fired:

safe_mutex: Trying to unlock mutex that wasn't locked at /home/seppo/work/wsrep/wsrep-5.6-UKnoPA/wsrep-5.6/sql/mysqld.cc, line 5505
            Last used at /home/seppo/work/wsrep/wsrep-5.6-UKnoPA/wsrep-5.6/include/mysql/psi/mysql_thread.h, line: 1162
06:36:47 UTC - mysqld got signal 6 ;

...

stack_bottom = 7fa5f4046e48 thread_stack 0x40000
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld(my_print_stacktrace+0x38)[0xadeec7]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld(handle_fatal_signal+0x3f3)[0x74543d]
/lib/x86_64-linux-gnu/libpthread.so.0(+0xfbb0)[0x7fa5f91d4bb0]
/lib/x86_64-linux-gnu/libc.so.6(gsignal+0x37)[0x7fa5f8615f77]
/lib/x86_64-linux-gnu/libc.so.6(abort+0x148)[0x7fa5f86195e8]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld(safe_mutex_unlock+0x8f)[0xae37e0]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld[0x615bf2]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld[0x6216ec]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld(_Z39wsrep_wait_committing_connections_closei+0xbc)[0x6217fc]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld[0x635e69]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld(_Z24wsrep_to_isolation_beginP3THDPcS1_PK10TABLE_LIST+0x3fe)[0x63682e]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld(_Z21mysql_execute_commandP3THD+0x21e4)[0x7fb53f]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld(_Z11mysql_parseP3THDPcjP12Parser_state+0x421)[0x8059ff]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld[0x804fee]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld(_Z16dispatch_command19enum_server_commandP3THDPcj+0x1067)[0x7f6e38]
2014-07-13 09:36:47 7832 [Note] WSREP: commit waiting for replaying: replayers 1, thd: (178) conflict: 0 (round: 1)
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld(_Z10do_commandP3THD+0x6c7)[0x7f593a]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld(_Z24do_handle_one_connectionP3THD+0x1be)[0x7bb863]
/home/seppo/work/testing/node1-shm-5.6/mysql/sbin/mysqld(handle_one_connection+0x33)[0x7bb366]
/lib/x86_64-linux-gnu/libpthread.so.0(+0x7f6e)[0x7fa5f91ccf6e]
/lib/x86_64-linux-gnu/libc.so.6(clone+0x6d)[0x7fa5f86d99cd]

[Seppo Jaakola (seppo-jaakola)]

Fix pushed for wsrep-5.6 in revision:
http://bazaar.launchpad.net/~codership/codership-mysql/5.6/revision/4108

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PXC-1232