Details
-
Bug
-
Status: Open (View Workflow)
-
Major
-
Resolution: Unresolved
-
10.1, 10.2, 10.3
Description
When the gtid-mode Slave receiver thread experiences ER_NET_READ_INTERRUPTED it carries out
a reconnect protocol that relay-logs an artificial Rotate event in front of events that follow the past break point.
It all looks like the Rotate causes a premature commit so the post-reconnect transferred Rows-events could not be handled:
#0 unlock_external (thd=, table=, count=2408550287) at lock.cc:720
|
#1 mysql_unlock_tables (thd=, sql_lock=, free_lock=false) at lock.cc:432
|
#2 mysql_unlock_tables (thd=, sql_lock=) at lock.cc:418
|
#3 close_thread_tables (thd=) at sql_base.cc:840
|
#4 rpl_group_info::slave_close_thread_tables (this=, thd=) at rpl_rli.cc:1939
|
#5 rpl_group_info::cleanup_context (this=, thd=, error=false) at rpl_rli.cc:1837
|
#6 rows_event_stmt_cleanup (rgi=, thd=) at log_event.cc:11614
|
#7 Rows_log_event::do_apply_event (this=, rgi=) at log_event.cc:11505
|
#8 Log_event::apply_event (this=, rgi=) at log_event.h:1482
|
#9 apply_event_and_update_pos_apply (ev=, thd=, rgi=, reason=0) at slave.cc:3579
|
#10 apply_event_and_update_pos (ev=, thd=, rgi=) at slave.cc:3727
|
#11 exec_relay_log_event (thd=, rli=, serial_rgi=) at slave.cc:4014
|
#12 handle_slave_sql (arg=) at slave.cc:5161
|
because of a table object was already destructed.
Attachments
Issue Links
- is duplicated by
-
MDEV-20095 MariaDB server replication crashing
- Open
- relates to
-
MDEV-20095 MariaDB server replication crashing
- Open
-
MDEV-24336 replica in crash loop on mariadb 10.5.7
- Closed