Re: Weird error - mysqld dump

From: B. Keith Murphy <kmurphy(at)icontact.com>
Date: Wed Oct 17 2007 - 01:24:30 EDT

Thanks for the reply Matthew. I should have outlined this before. The file table was on an ndb engine and it is part of the mogile application. Other than the mysql table everything on this cluster is running on ndb tables. Here is the file table:

mysql> describe file;

+----------+----------------------+------+-----+---------+-------+ 

| Field | Type | Null | Key | Default | Extra |

+----------+----------------------+------+-----+---------+-------+ 
| fid | int(10) unsigned | NO | PRI | | | 
| dmid | smallint(5) unsigned | NO | MUL | | | 
| dkey | varchar(255) | YES | | NULL | | 
| length | int(10) unsigned | YES | | NULL | | 

| classid | tinyint(3) unsigned | NO | | | | | devcount | tinyint(3) unsigned | NO | | | |

+----------+----------------------+------+-----+---------+-------+ 

6 rows in set (0.00 sec)

mysql>

It currently contains right at 7,500 entries.

It dumped again this afternoon. That is the third time..actually just within the 24 hour time frame. Same symptoms. It is always dumps on a REPLACE INTO statement. This statement (with various values) is executed quite frequently though so it can't be a syntax problem. Another words, it might execute 1,000 times and only crash once. Hope that makes sense.

I just finished upgrading to 5.1.22. I doubt it will make any difference but I had to do something :).

The ndbd nodes seem to be unaffected. The error logs for them don't show any problems when this is happening. Since this is a development cluster we run two sql nodes and two data nodes. One sql node and one data node each run on a server (two servers total).

thanks for the help,

Keith

• Original Message ----- From: "Matthew Montgomery" <mmontgomery@mysql.com> To: "B. Keith Murphy" <kmurphy@icontact.com> Cc: "cluster" <cluster@lists.mysql.com> Sent: Tuesday, October 16, 2007 3:51:00 PM (GMT-0500) America/New_York Subject: Re: Weird error - mysqld dump

Keith,

On Tue, 2007-10-16 at 11:26 -0400, B. Keith Murphy wrote:
> I have a problem with a test cluster (5.1.20) I set up. We are beginning
> to run tests against it and twice in the last 24 hours one of the sql
> nodes has dumped. The daemon.log file shows the same command is being executed:
> Oct 15 16:06:19 dev-db1c mysqld[18358]: Trying to get some variables.
> Oct 15 16:06:19 dev-db1c mysqld[18358]: Some pointers may be invalid and cause the dump to abort...
> Oct 15 16:06:19 dev-db1c mysqld[18358]: thd->query at 0x8e04c88 = REPLACE
> INTO file (fid, dmid, dkey, length, classid, devcount) VALUES ('166970','4','Post_412_Content','12','0',0)
> Oct 15 16:06:19 dev-db1c mysqld[18358]: thd->thread_id=9

> Both times it dumped the same basic statement was being executed. I
> have looked at the obvious things..running out of memory, not enough
> resources allocated, etc. I am not coming up with anything. This
> replace statement is trying to insert into a MogileFS database by the
> way. This is pretty much a shot in the dark but I am hoping someone
> can help :)
>

I don't have much experience with MogileFS, but I'll try to respond to your questions with a few of my own. I assume that `file` is a part of the MogileFS tracker application. What engine is the `file` table stored in? What is the schema of that table? How many rows are in that table? If the table is in ndbcluster, do you notice any ndbd node failures at the time of the mysqld crash?

thanks,

Keith

-- 
Matthew Montgomery, Support Engineer 
San Antonio, TX 
MySQL Inc, www.mysql.com 



-- 
B. Keith Murphy 
Database Administrator 
iContact 
2635 Meridian Parkway, 2nd Floor 
Durham, North Carolina 27713 
blog: 
http://blog.paragon-cs.com 
(o) 919-433-0786 
(c) 850-637-3877 

Received on Wed Oct 17 01:27:03 2007