LP:625724 - Aria recovery assertion ma_recovery.c:2959: get_MARIA_HA_from_REDO_record: Assertion `((rec->lsn) - (checkpoint_start)) < 0'

Update by Elena on 2012-11-01:

I could not reproduce the problem using the RQG workload from the initial description, but the assertion itself is still reproducible by trying to start the server on the attached datadir.

So, to get the assertion, I don't recommend spending time trying to run the RQG test. Instead,

• download and unzip the attached zip archive, it will create var-bug625724 folder;
• start debug MySQL server with --datadir=<path>/var-bug625724/master-data

It takes time, about 5-10 min, but then fails with the described assertion.

I tried to do the same on a release (non-debug) server, nothing bad seems to happen, the server starts and works, no complaints in the log, which is why I'm setting the priority to Minor for now.

Reproducible on maria/5.1 revno 3151, maria/5.3 revno 3589, maria/10.0 revno 3467. I didn't check maria/5.2, but presumably it shouldn't be any different there.

==============================
Initial description:

When recovering from a DML workload, aria asserted as follows:

recovered pages: 0% 10% 20%mysqld: ma_recovery.c:2959: get_MARIA_HA_from_REDO_record: Assertion `((rec->lsn) - (checkpoint_start)) < 0' failed.

rqg command line :

 perl runall.pl --basedir=/home/philips/bzr/maria-5.2/ --grammar=conf/smf/smf2.yy --mysqld=--init_file=/home/philips/bzr/randgen-smf/conf/smf/smf2.sql --queries=10M --duration=360 --engine=Maria --mysqld=--skip-innodb --mysqld=--default-storage-engine=Maria --duration=600 --mysqld=--log-output=file --mem --duration=3600 --reporter=Recovery --mysqld=--maria-checkpoint-interval=0

$ bzr version-info

revision-id: <email address hidden>

date: 2010-08-25 01:47:52 +0300

build-date: 2010-08-28 11:54:04 +0300

revno: 2848

branch-nick: maria-5.2