Second ma_blockrec.c:3000: write_block_record: Assertion `cur_block[1].page_count == 0' failed with a multi-index Aria workload

[Philip Stoev]

When executing the many_indexes workload, Aria asserted as follows:

mysqld: ma_blockrec.c:3000: write_block_record: Assertion `cur_block[1].page_count == 0' failed.

# 2011-07-22T20:56:21 #6 0x00000033b58296e6 in __assert_fail () from /lib64/libc.so.6
# 2011-07-22T20:56:21 #7 0x0000000000980ad9 in write_block_record (info=0x2aaab8c01748, old_record=0x2aaab8bfb530 "", record=0x2aaab8bfa0f0 "", row=0x2aaab8c
01868,
# 2011-07-22T20:56:21 bitmap_blocks=0x2aaab8c017a8, head_block_is_read=1 '\001', row_pos=0x40a04150, undo_lsn=1, old_record_checksum=0) at ma_blockrec.c:3
000
# 2011-07-22T20:56:21 #8 0x0000000000982c04 in _ma_update_block_record2 (info=0x2aaab8c01748, record_pos=3072, oldrec=0x2aaab8bfb530 "", record=0x2aaab8bfa0f
0 "", undo_lsn=1)
# 2011-07-22T20:56:21 at ma_blockrec.c:3809
# 2011-07-22T20:56:21 #9 0x00000000009832d2 in _ma_update_block_record (info=0x2aaab8c01748, record_pos=3072, orig_rec=0x2aaab8bfb530 "", new_rec=0x2aaab8bfa
0f0 "")
# 2011-07-22T20:56:21 at ma_blockrec.c:3973
# 2011-07-22T20:56:21 #10 0x0000000000973817 in maria_update (info=0x2aaab8c01748, oldrec=0x2aaab8bfb530 "", newrec=0x2aaab8bfa0f0 "") at ma_update.c:161
# 2011-07-22T20:56:21 #11 0x000000000093632b in ha_maria::update_row (this=0x2aaab8bf98f0, old_data=0x2aaab8bfb530 "", new_data=0x2aaab8bfa0f0 "") at ha_maria
.cc:2135
# 2011-07-22T20:56:21 #12 0x00000000007c19d1 in handler::ha_update_row (this=0x2aaab8bf98f0, old_data=0x2aaab8bfb530 "", new_data=0x2aaab8bfa0f0 "") at handle
r.cc:4927
# 2011-07-22T20:56:21 #13 0x00000000007d07c2 in ha_partition::update_row (this=0x2aaab8bf8cf0, old_data=0x2aaab8bfb530 "", new_data=0x2aaab8bfa0f0 "") at ha_p
artition.cc:3437
# 2011-07-22T20:56:21 #14 0x00000000007c19d1 in handler::ha_update_row (this=0x2aaab8bf8cf0, old_data=0x2aaab8bfb530 "", new_data=0x2aaab8bfa0f0 "") at handle
r.cc:4927
# 2011-07-22T20:56:21 #15 0x0000000000739b31 in mysql_update (thd=0x2aaab8b67798, table_list=0x1a5c2420, fields=..., values=..., conds=0x1a5c2cd0, order_num=1
, order=0x1a5c2e88,
# 2011-07-22T20:56:21 limit=1, handle_duplicates=DUP_ERROR, ignore=false) at sql_update.cc:674
# 2011-07-22T20:56:21 #16 0x000000000067d04b in mysql_execute_command (thd=0x2aaab8b67798) at sql_parse.cc:3110
# 2011-07-22T20:56:21 #17 0x00000000006838a1 in mysql_parse (thd=0x2aaab8b67798,
# 2011-07-22T20:56:21 rawbuf=0x1a5c2170 "UPDATE `table10_aria_key_pk_parts_2_int_autoinc` SET `col_varchar_255_ucs2_key` = REPEAT( _UTF8 0x7 , 38 ) WHERE
`col_longtext_ucs2` IS NOT NULL ORDER BY `pk` LIMIT 9", length=166, found_semicolon=0x40a05f18) at sql_parse.cc:6138
# 2011-07-22T20:56:21 #18 0x0000000000684732 in dispatch_command (command=COM_QUERY, thd=0x2aaab8b67798,
# 2011-07-22T20:56:21 packet=0x2aaab8be17b9 "UPDATE `table10_aria_key_pk_parts_2_int_autoinc` SET `col_varchar_255_ucs2_key` = REPEAT( _UTF8 0x7 , 38 ) WH
ERE `col_longtext_ucs2` IS NOT NULL ORDER BY `pk` LIMIT 9", packet_length=166) at sql_parse.cc:1213
# 2011-07-22T20:56:21 #19 0x0000000000685d4f in do_command (thd=0x2aaab8b67798) at sql_parse.cc:903
# 2011-07-22T20:56:21 #20 0x00000000006708c8 in handle_one_connection (arg=0x2aaab8b67798) at sql_connect.cc:1177
# 2011-07-22T20:56:21 #21 0x00000033b600673d in start_thread () from /lib64/libpthread.so.0
# 2011-07-22T20:56:21 #22 0x00000033b58d40cd in clone () from /lib64/libc.so.6
# 2011-07-22T20:56:21 #5 0x00000033b5831d10 in abort () from /lib64/libc.so.6

RQG command lines:

runall.pl --queries=100000000 --queries=1M --engine=Aria --mysqld=--default-storage-engine=Aria --mysqld=--safe-mode --mysqld=--loose-debug-assert-if-crashed-table --mysqld=--sync-sys=0 --mysqld=--log-output=file --mysqld=--aria_log_purge_type=at_flush --reporters=ErrorLog,Backtrace,AriaDoubleRecovery,Shutdown --duration=240 --threads=5 --rows=10 --mask-level=1 --mysqld=--aria-repair-threads=2 --mysqld=--loose-aria-group-commit=hard --mysqld=--loose-aria_group_commit_interval=1 --mysqld=--aria-checkpoint-interval=0 --mysqld=--aria-block-size=1K --mysqld=--table_cache=32K --mysqld=--aria-pagecache-division-limit=75 --grammar=conf/engines/many_indexes.yy --gendata=conf/engines/many_indexes.zz --mtr-build-thread=300 --mask=24395 --basedir1=/home/philips/bzr/maria-5.2 --seed=1311358797

runall.pl --queries=100000000 --queries=1M --engine=Aria --mysqld=--default-storage-engine=Aria --mysqld=--safe-mode --mysqld=--loose-debug-assert-if-crashed-table --mysqld=--sync-sys=0 --mysqld=--log-output=file --mysqld=--aria_log_purge_type=at_flush --reporters=ErrorLog,Backtrace,AriaDoubleRecovery,Shutdown --duration=120 --threads=10 --rows=10 --mask-level=1 --mysqld=--loose-aria-group-commit=soft --mysqld=--loose-aria_group_commit_interval=10 --mysqld=--aria-checkpoint-interval=0 --mysqld=--aria-block-size=2K --mysqld=--aria-pagecache-division-limit=75 --grammar=conf/engines/many_indexes.yy --gendata=conf/engines/many_indexes.zz --mtr-build-thread=300 --mask=21095 --basedir1=/home/philips/bzr/maria-5.2 --seed=1311344732

runall.pl --queries=100000000 --queries=1M --engine=Aria --mysqld=--default-storage-engine=Aria --mysqld=--safe-mode --mysqld=--loose-debug-assert-if-crashed-table --mysqld=--sync-sys=0 --mysqld=--log-output=file --mysqld=--aria_log_purge_type=at_flush --reporters=ErrorLog,Backtrace,AriaDoubleRecovery,Shutdown --duration=480 --threads=20 --rows=10 --mask-level=2 --mysqld=--loose-aria-group-commit=hard --mysqld=--loose-aria_group_commit_interval=100 --mysqld=--aria-checkpoint-interval=120 --mysqld=--aria-block-size=1K --mysqld=--aria-pagecache-buffer-size=512K --grammar=conf/engines/many_indexes.yy --gendata=conf/engines/many_indexes.zz --mtr-build-thread=300 --mask=54704 --basedir1=/home/philips/bzr/maria-5.2 --seed=1311357374

bzr version-info

revision-id: <email address hidden>
date: 2011-07-21 15:21:22 +0300
build-date: 2011-07-23 13:05:07 +0300
revno: 3004
branch-nick: maria-5.2

[Philip Stoev (pstoev-askmonty)]

See also bug 726374

[Michael Widenius (monty)]

Great test case; Found 3 unlikely but serious bugs in the ARIA code thanks to it!

I have now run all the above 3 test for some period without failures, so I think this is now fixed.
I will push the fixes into 5.2