Memory engine table full at much less than max_heap_table_size with btree index

[Jason Preuss]

This is the same bug as listed on Mysql http://bugs.mysql.com/bug.php?id=44138. I am confirming that the bug is also present in mariadb-5.1.51-5.1.x86_64. Let me know if you need anything in addition to what is already listed at the Mysql site.

[Jason Preuss (preuss)]

It has been 3 months and it has yet to be assigned, any hope someone will take a look at this?

[Philip Stoev (pstoev-askmonty)]

Assigning to Monty, as he may be working in this area of the code shortly.

[Michael Widenius (monty)]

I am now working on a new memory engine; I will check this issue at the same time.
ETA: Sometimes this month (October)

[Jason Preuss (preuss)]

Any update on the time frame?

[Michael Widenius (monty)]

Sorry for delay. We have been busy with getting 5.3 and 5.5 out.
I asked holyfoot, who is one of the authors of the btree indexes in the memory engine to take a look at this.

[Alexey Botchkov (holyfoot)]

I did a lot of tests with various maria/mysql versions on one of our machines:
Linux terrier 2.6.34.8-0.2-desktop #1 SMP PREEMPT 2011-04-06 18:11:26 +0200 x86_64 x86_64 x86_64 GNU/Linux

All of them handled that b129_SNPMapInfo_36_3.bcp.gz dataset with no failures.

So it seems i need some help to understand the problem.
Jason, can you recommend anything?

[Jason Preuss (preuss)]

Alexey,

Thanks for looking into this.

The b129_SNPMapInfo_36_3.bcp.gz dataset was from the person who first reported this issue to MySQL nearly 3 years ago and was not the data I was having problems with. I did, however, this morning try to replicate his problem from his data and instructions and I got the exact same behavior as he did. I have attached a file of my output, system specs, and my.conf for my system, I hope that will be useful.

I am not sure what else to say. 3 different people on different hardware and linux distros have run into this behavior. I personally have tested mysql5.0, 5.1, 5.5 and mariadb5.1 on SLES 10, 11, and Ubuntu 10.04 and get the exact same behavior every time so I'm pretty sure there is a problem. If there is any other information I can provide please let me know, solving this problem will save me countless hours of processing time.

[Alexey Botchkov (holyfoot)]

Jason,

i did my best to try to reproduce this problem on the machines i have at my hand, but wasn't able to do it.
Can i ask you for an assistance?
An access to one of the machines where the problem manifests would be perfect. If it's not possible i'd ask you to collect --debug log somewhere there.

[Jason Preuss (preuss)]

Alexey,

A debug file should be no problem but I am not the head administrator of the machine so I can't promise access . I do fully understand that it is close to impossible to fix a problem you can't replicate and that having access to a machine with the problem is the way to go. I am not in the office today so I will get back to you on Monday.

Thanks again for all your work on this.

Jason

[Alexey Botchkov (holyfoot)]

pushed into maria/5.1 and above.