Description:
Occasionally the mysqld node crashes with a segfault when a database containing NDB tables is dumped. The crash happens early, before any data is retrieved:
mysqldump: mysqldump: Couldn't execute 'show table status like 'file\_msg'': Lost connection to MySQL server during query (2013)
mysqldump: mysqldump: Couldn't execute 'SET OPTION SQL_QUOTE_SHOW_CREATE=1': MySQL server has gone away (2006)
mysqldump: mysqldump: Couldn't execute 'show fields from `file_msg`': MySQL server has gone away (2006)
The MySQL instance contains both MyISAM and NDB tables and is dumped schema by schema. The crash happens for different schemas but always immediately after stating mysqldump.
Two core files show the same stack trace:
#0 0x0000003abc80b132 in pthread_kill () from /lib64/libpthread.so.0
#1 0x0000000000589a76 in handle_segfault ()
#2 <signal handler called>
#3 0x00000000006b66fa in Ndb::readAutoIncrementValue ()
#4 0x0000000000649afa in ha_ndbcluster::info ()
#5 0x00000000006671be in get_schema_tables_record ()
#6 0x00000000006668c3 in get_all_tables ()
#7 0x000000000066bdf8 in get_schema_tables_result ()
#8 0x00000000005d4d7b in JOIN::exec ()
#9 0x00000000005d50ea in mysql_select ()
#10 0x00000000005d0ade in handle_select ()
#11 0x000000000059f75e in mysql_execute_command ()
#12 0x00000000005a588c in mysql_parse ()
#13 0x000000000059df1e in dispatch_command ()
#14 0x00000000005a91ff in do_command ()
#15 0x000000000059d205 in handle_one_connection ()
#16 0x0000003abc8062f7 in start_thread () from /lib64/libpthread.so.0
#17 0x0000003abb4ce85d in clone () from /lib64/libc.so.6
How to repeat:
analyzing
