Aria recovery: ma_blockrec.c:3930: _ma_update_at_original_place: Assertion `block->org_bitmap_value == _ma_bitmap_get_page_bits(info, &info->s->bitmap, page)' failed

[Philip Stoev]

When recovering from the SMF scenario with maria_read_log, the recovery asserted as follows:

transactions to roll back: 2maria_read_log: ma_blockrec.c:3930: _ma_update_at_original_place: Assertion `block->org_bitmap_value == _ma_bitmap_get_page_bits(info, &info->s->bitmap, page)' failed.

bzr version-info:

revno: 3006
committer: Michael Widenius <email address hidden>
branch nick: maria-5.1
timestamp: Fri 2011-01-07 17:58:12 +0200
message:
  Don't do DBUG_ASSERT for checksum errors when using REPAIR
  mysql_convert_table_format ignored --engine option.
  Fix that zerofill() doesn't write out wrong data to client if run with auto repair.
  Ensure that pagecache is properly flushed, even in case of errors.
  Handle checksum errors in BLOCK_RECORD format.

RQG command line:

perl runall.pl \
--queries=1M \
--engine=Maria \
--mysqld=--default-storage-engine=Maria \
--mysqld=--safe-mode \
--mysqld=--loose-debug-assert-if-crashed-table \
--mysqld=--sync-sys=0 \
--mysqld=--log-output=file \
--mysqld=--maria_log_purge_type=at_flush \
--reporters=ErrorLog,Backtrace,Recovery,Shutdown \
--duration=240 \
--threads=10 \
--rows=10 \
--mask-level=2 \
--mysqld=--maria-repair-threads=2 \
--mysqld=--loose-maria-group-commit=hard \
--mysqld=--loose-maria_group_commit_interval=10 \
--mysqld=--maria-checkpoint-interval=120 \
--mysqld=--maria-block-size=16K \
--queries=100000000 \
--mask=735 \
--seed=1294540351\
--mysqld=--init-file=/home/buildbot/randgen/conf/smf/smf2.sql \
--grammar=conf/smf/smf2.yy \
--basedir=.

[Philip Stoev (philip-stoev)]

vardir before recovery, log, etc.

[Michael Widenius (monty)]

Found and fixed. Will push tomorrow (just want to run the test for 12+ hours before committing).

[Michael Widenius (monty)]

Found another case where we got the same assert. Fixed last night and pushed.