MySQL Bugs: #51408: MyISAM crash during enable_indexes at unlink

Bug #51408	MyISAM crash during enable_indexes at unlink_block
Submitted:	23 Feb 2010 3:20	Modified:	1 Jul 2010 18:41
Reporter:	James Day	Email Updates:
Status:	No Feedback	Impact on me:	None
Category:	MySQL Server: MyISAM storage engine	Severity:	S2 (Serious)
Version:	5.0.60sp1	OS:	Linux (2.6.18-164.el5 #1 SMP)
Assigned to:		CPU Architecture:	Any

Description:
Server crash while MyISAM is doing a repair after bulk insert.

0x81a0705 handle_segfault + 805
0x8407eef unlink_block + 15
0x8408a8f free_block + 159
0x8408b75 flush_cached_blocks + 149
0x8408dd2 flush_key_blocks_int + 354
0x8409af7 flush_key_blocks + 55
0x83e7fc9 flush_blocks + 41
0x83eefc3 mi_repair_by_sort + 451
0x826014c ha_myisam::repair(THD*, st_mi_check_param&, bool) + 1900
0x826035c ha_myisam::enable_indexes(unsigned int) + 364

A similar stack trace is in bug #22879. I wonder if bug #34634 might be related?

How to repeat:
Not yet known.

I get a similar crash in 5.1.45 using sysbench sbtest table. The basic reproduction case is:
1) create table without indexes
2) insert 40M rows from another table
3) alter table add indexes
4) get disk full errors
4) hit control-C

error log output:
100318 18:13:23 [ERROR] /data/5145fb/libexec/mysqld: The table 'sbtest' is full
100318 19:09:02 [ERROR] /data/5145fb/libexec/mysqld: Disk is full writing './testm/#sql-452b_df.MYD' (Errcode: 28). Waiting for someone to 
free space... (Expect up to 60 secs delay for server to continue after freeing disk space)
100318 19:09:02 [ERROR] /data/5145fb/libexec/mysqld: Retry in 60 secs. Message reprinted in 600 secs
100318 19:13:02 [ERROR] /data/5145fb/libexec/mysqld: Incorrect key file for table './testm/#sql-452b_df.MYI'; try to repair it

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 = 0x41b150e0 thread_stack 0x40000
/data/5145fb/libexec/mysqld(my_print_stacktrace+0x2e) [0x87953e]
/data/5145fb/libexec/mysqld(handle_segfault+0x31d) [0x5973fd]
/lib64/libpthread.so.0 [0x34bdc0de70]
/data/5145fb/libexec/mysqld [0x879f00]
/data/5145fb/libexec/mysqld [0x879f82]
/data/5145fb/libexec/mysqld [0x87ac67]
/data/5145fb/libexec/mysqld(flush_key_blocks+0x6f) [0x87c7df]
/data/5145fb/libexec/mysqld(flush_blocks+0x25) [0x7fb5e5]
/data/5145fb/libexec/mysqld(mi_repair_by_sort+0x162) [0x7fca22]
/data/5145fb/libexec/mysqld(ha_myisam::repair(THD*, st_mi_check_param&, bool)+0x6cf) [0x7dd90f]
/data/5145fb/libexec/mysqld(ha_myisam::enable_indexes(unsigned int)+0x114) [0x7dda94]
/data/5145fb/libexec/mysqld(ha_myisam::end_bulk_insert()+0x61) [0x7defd1]
/data/5145fb/libexec/mysqld(mysql_alter_table(THD*, char*, char*, st_ha_create_information*, TABLE_LIST*, Alter_info*, unsigned int, st_ord
er*, bool)+0x2356) [0x683856]
/data/5145fb/libexec/mysqld(mysql_execute_command(THD*, unsigned long long*)+0x28ae) [0x5a70ce]
/data/5145fb/libexec/mysqld(mysql_parse(THD*, char const*, unsigned int, char const**, unsigned long long*)+0x239) [0x5aa5d9]
/data/5145fb/libexec/mysqld(dispatch_command(enum_server_command, THD*, char*, unsigned int)+0x45b) [0x5aaa7b]
/data/5145fb/libexec/mysqld(do_command(THD*)+0xe7) [0x5abbe7]
/data/5145fb/libexec/mysqld(handle_one_connection+0x55f) [0x59e37f]
/lib64/libpthread.so.0 [0x34bdc062f7]
/lib64/libc.so.6(clone+0x6d) [0x34bd0d1e3d]
Trying to get some variables.
Some pointers may be invalid and cause the dump to abort...
thd->query at 0xcb004c0 = alter table sbtest add primary key(id), add index k(k)
thd->thread_id=223
thd->killed=KILL_QUERY

my.cnf settings:
[mysqld]
plugin-load=libpbxt.so

pbxt
pbxt_index_cache_size=250M
pbxt_record_cache_size=750M

innodb_buffer_pool_size=1000M
innodb_log_file_size=100M
innodb_flush_log_at_trx_commit=2
innodb_doublewrite=0
# innodb_flush_method=O_DIRECT
innodb_thread_concurrency=0
max_connections=2000
table_cache=2000
innodb_max_dirty_pages_pct=80

key_buffer_size=1000M

table DDL before adding indexes:
| sbtest | CREATE TABLE `sbtest` (
  `id` int(10) unsigned NOT NULL DEFAULT '0',
  `k` int(10) unsigned NOT NULL DEFAULT '0',
  `c` char(120) NOT NULL DEFAULT '',
  `pad` char(60) NOT NULL DEFAULT ''
) ENGINE=MyISAM DEFAULT CHARSET=latin1 |

alter table statement
alter table sbtest add primary key(id), add index k(k);

Mark,

thank you for the feedback. Could you please try 5.1.47 in your environment: I was not able to repeat the problem using this version.

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".