RQG: Assertion "crc" == "failure in REDO_INDEX" on Maria engine recovery

[Philip Stoev]

When recovering from a kill -9 crash that happened during a DML workload, maria 5.3 crashed as follows:

#8 0x00897de8 in __assert_fail () from /lib/libc.so.6
#9 0x08557165 in _ma_apply_redo_index (info=0x9c94d60, lsn=4297618642, header=0x9c75816 "\346\017\271*\267\273", head_length=18) at ma_key_recover.c:987
#10 0x085bda51 in exec_REDO_LOGREC_REDO_INDEX (rec=0xbfe2a8b0) at ma_recovery.c:1571
#11 0x085bb5db in display_and_apply_record (log_desc=0x9265440, rec=0xbfe2a8b0) at ma_recovery.c:517
#12 0x085bfee3 in run_redo_phase (lsn=4294975495, apply=MARIA_LOG_APPLY) at ma_recovery.c:2507
#13 0x085baf94 in maria_apply_log (from_lsn=4294975495, apply=MARIA_LOG_APPLY, trace_file=0x9c251a0, should_run_undo_phase=1 '\001', skip_DDLs_arg=1 '\001',
    take_checkpoints=1 '\001', warnings_count=0xbfe2d434) at ma_recovery.c:316
#14 0x085bac7c in maria_recovery_from_log () at ma_recovery.c:218
#15 0x085404f3 in ha_maria_init (p=0x9c16dd0) at ha_maria.cc:3265
#16 0x083b9d09 in ha_initialize_handlerton (plugin=0x9c08f08) at handler.cc:429
#17 0x08473e5f in plugin_initialize (plugin=0x9c08f08) at sql_plugin.cc:1038
#18 0x084746d5 in plugin_init (argc=0x8a4e318, argv=0x9bdba40, flags=0) at sql_plugin.cc:1267
#19 0x0827a0ec in init_server_components () at mysqld.cc:4113
#20 0x0827abb2 in main (argc=14, argv=0xbfe2da24) at mysqld.cc:4587

The tablespaces before and after recovery will be uploaded shortly.

[Philip Stoev (pstoev-askmonty)]

[philips@eve mysql-test]$ bzr version-info
revision-id: <email address hidden>
date: 2010-06-14 15:17:54 +0400
build-date: 2010-07-06 01:21:22 -0700
revno: 2789
branch-nick: maria

[Philip Stoev (pstoev-askmonty)]

Please ignore last attachment, it was for a different bug. Please use the first , original attachment when debugging this bug.

[Philip Stoev (pstoev-askmonty)]

RQG command line:

# 2010-08-25T02:11:48 # /home/buildbot/randgen/runall.pl \
# 2010-08-25T02:11:48 # --basedir=. \
# 2010-08-25T02:11:48 # --vardir=../../vardir-maria_varchar \
# 2010-08-25T02:11:48 # --grammar=/home/buildbot/randgen/conf/engines/varchar.yy \
# 2010-08-25T02:11:48 # --gendata=/home/buildbot/randgen/conf/engines/varchar.zz \
# 2010-08-25T02:11:48 # --reporter=Backtrace,ErrorLog,Recovery,Shutdown \
# 2010-08-25T02:11:48 # --duration=240 \
# 2010-08-25T02:11:48 # --queries=1M \
# 2010-08-25T02:11:48 # --engine=Maria \
# 2010-08-25T02:11:48 # --mysqld=--maria-checkpoint-interval=0 \
# 2010-08-25T02:11:48 # --mysqld=--log-output=file \
# 2010-08-25T02:11:48 # --seed=time \
# 2010-08-25T02:11:48 # --mysqld=--loose-skip-innodb \
# 2010-08-25T02:11:48 # --mysqld=--loose-pbxt=OFF

This bug is readily repeatable in buildbot on the varchar.yy test.