mysqld crash - audit_plugin.cc:665

[outcassed]

mysqld crash - audit_plugin.cc:665

[glicht]

Thanks for reporting this. Much appreciated.

Can you provide some info about the mysql configuration. What mysql distribution/version and what are the settings you have for the audit plugin (show global variables like '%audit%').

The bug looks very similar to issue #29. It might be that the fix for #29 didn't resolve the issue fully. At that case I think the node was used at replication. In your case, is this node used in replication by any chance? If so please provide some info about the setup.

[outcassed]

Hi there - this was a master in a single master setup.

I was using:

audit_json_file=0
audit_json_socket_name=/dev/mysqlaudit
audit_json_socket=1
audit_record_objs=database.table1,database.table2

The crash didn't occur immediately. Auditing was working properly, and about a day later the crash occurred.

I'm running Percona-Server-5.5.11-rel20.2 (based on MySQL 5.5.11)

[outcassed]

I should add - it may have been a coincidence but the crash seemed to occur just as I was launching a MySQL GUI tool (Sequel Pro) The previous time that it crashed, it was under normal traffic and there weren't any admin tools/DDL statements/etc.

The commands that Sequel Pro issues upon launch are:

SHOW VARIABLES;
SELECT @@global.max_allowed_packet;
USE database_name;
SHOW DATABASES;
SHOW /!50002 FULL/ TABLES;
SELECT * FROM information_schema.routines WHERE routine_schema = 'database_name' ORDER BY routine_name;
SHOW VARIABLES LIKE 'character_set_database';

I don't believe that I focused on one of the audited tables before the crash, but if I had, it would have issued:

SHOW TABLE STATUS LIKE 'table_name';
SHOW CREATE TABLE table_name;
SELECT COUNT(1) FROM table_name;

[glicht]

Do you mean by "single master" that there are no slaves connecting?

---

Sent from my mobile device
On Feb 2, 2013 5:47 PM, "Casey Forbes" [email protected] wrote:

Hi there - this was a master in a single master setup.

I was using:

audit_json_file=0
audit_json_socket_name=/dev/mysqlaudit
audit_json_socket=1
audit_record_objs=database.table1,database.table2

The crash didn't occur immediately. Auditing was working properly, and
about a day later the crash occurred.

I'm running Percona-Server-5.5.11-rel20.2 (based on MySQL 5.5.11)

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/31#issuecomment-13032352.

[outcassed]

Sorry, I don't know what I meant by that :) Yes, there are slaves connected. I'm only using the plugin on the master.

[glicht]

Just to update. I think I have a scenario regarding this crash. It is associated to how we check objects for non query commands. There could be some scenarios that the query tables list is pointing to garbage. Will be working on putting some safe guards to avoid this.

[glicht]

Latest commit should fix the crash issue. If you still experience the crash please re-open the issue.

[outcassed]

Great!