Skip to:

  1. Skip to Jira Navigation
  2. Skip to Side Navigation
  3. Skip to Main Content

PXC crash when updating same rows in multiple nodes

Description

PXC crashes with "mysqld got signal 11" when updating same rows in multiple nodes concurrently. It seems the crash happens while checking for lock.

Thread pointer: 0x7f36940008c0 Attempting backtrace. You can use the following information to find out where mysqld died. If you see no messages after this, something went terribly wrong... stack_bottom = 7f36a92add50 thread_stack 0x40000 /usr/sbin/mysqld(my_print_stacktrace+0x3b)[0xf416fb] /usr/sbin/mysqld(handle_fatal_signal+0x471)[0x7b5c51] /lib64/libpthread.so.0(+0xf5d0)[0x7f36c08715d0] /usr/sbin/mysqld(my_free+0x18)[0xf3dba8] /usr/sbin/mysqld(_ZN14Relay_log_info20clear_tables_to_lockEv+0x7b)[0xf0b23b] /usr/sbin/mysqld(_ZN14Relay_log_info15cleanup_contextEP3THDb+0x8b)[0xf0b3ab] /usr/sbin/mysqld(_ZN14Rows_log_event14do_apply_eventEPK14Relay_log_info+0x7e8)[0xeb3c98] /usr/sbin/mysqld(_ZN9Log_event11apply_eventEP14Relay_log_info+0x6d)[0xeab97d] /usr/sbin/mysqld(_Z14wsrep_apply_cbPvPKvmjPK14wsrep_trx_meta+0x65a)[0x7d045a] /usr/lib64/galera3/libgalera_smm.so(_ZNK6galera9TrxHandle5applyEPvPF15wsrep_cb_statusS1_PKvmjPK14wsrep_trx_metaERS6_+0xfd)[0x7f36b45d04fd] /usr/lib64/galera3/libgalera_smm.so(+0x1f47e3)[0x7f36b460f7e3] /usr/lib64/galera3/libgalera_smm.so(_ZN6galera13ReplicatorSMM9apply_trxEPvPNS_9TrxHandleE+0x199)[0x7f36b4612419] /usr/lib64/galera3/libgalera_smm.so(_ZN6galera13ReplicatorSMM11process_trxEPvPNS_9TrxHandleE+0x14e)[0x7f36b461552e] /usr/lib64/galera3/libgalera_smm.so(_ZN6galera15GcsActionSource8dispatchEPvRK10gcs_actionRb+0x1d0)[0x7f36b45efd90] /usr/lib64/galera3/libgalera_smm.so(_ZN6galera15GcsActionSource7processEPvRb+0x66)[0x7f36b45f14f6] /usr/lib64/galera3/libgalera_smm.so(_ZN6galera13ReplicatorSMM10async_recvEPv+0x8b)[0x7f36b4615d6b] /usr/lib64/galera3/libgalera_smm.so(galera_recv+0x2c)[0x7f36b462801c] /usr/sbin/mysqld[0x7d192a] /usr/sbin/mysqld(start_wsrep_THD+0x222)[0x7a6782] /usr/sbin/mysqld(pfs_spawn_thread+0x1b4)[0xf59ff4] /lib64/libpthread.so.0(+0x7dd5)[0x7f36c0869dd5] /lib64/libc.so.6(clone+0x6d)[0x7f36bea3802d]Trying to get some variables. Some pointers may be invalid and cause the dump to abort. Query (0): is an invalid pointer Connection ID (thread ID): 1 Status: NOT_KILLED

 

How to reproduce:

Install 3 PXC nodes and run the bash script attached in multiple nodes simultaneously. To improve the change of the issue is recommended to run more than one instance of the bash script per node.

Please also find the SQL file with the schema and initial data and also the MySQL configuration file used in the tests.

Environment

None

AFFECTED CS IDs

263259

Attachments

1

Smart Checklist

Activity

Show:

Ramesh Sivaraman September 17, 2019 at 9:07 AM
Edited

Could not reproduce the issue with .7.27 package using the given test case.  The testcase from is also not crashed the server.

KennT August 26, 2019 at 1:54 AM

This bug is related to , which will be fixed by the merge with PS 5.7.27

KennT August 18, 2019 at 1:05 AM

I have repro-ed this with the latest 5.7.26 code.  It occurs when using log_slow_verbosity=innodb (also 'full' but that includes 'innodb').  Although I see an error during rollback.

Done

Details

Assignee

Reporter

Time tracking

2d 7h logged

Fix versions

Affects versions

Priority

Smart Checklist

Created August 15, 2019 at 7:28 AM
Updated March 6, 2024 at 10:07 PM
Resolved September 18, 2019 at 6:12 AM
Configure