MySQL down – again
MySQL server crashed again this morning:
Version: '4.1.14-log' socket: '' port: 3306 Source distribution
071203 7:35:45 InnoDB: Error: the OS said file flush did not succeed
071203 7:35:45 InnoDB: Operating system error number 12 in a file operation.
InnoDB: Error number 12 means 'not enough core'.
InnoDB: Some operating system error numbers are described at
InnoDB: http://dev.mysql.com/doc/mysql/en/Operating_System_error_codes.html
InnoDB: File operation call: 'flush'.
InnoDB: Cannot continue operation.
%DELETE-W-SEARCHFAIL, error searching for MYSQL_ROOT:[MYSQL_SERVER.TMP]*.*;*
-RMS-E-FNF, file not found
MYSQL_SERVER job terminated at 3-DEC-2007 07:35:47.95
The watcher wasn’t running, but since logging ws disabled, There is no way to locate the reason. The last log is of 26-Nov-2007, 23:32 so it might be that changing data may have caused an error. But even in acconting, it cannot be found.
So I just started it again – with logging enabled, to find out – eventually.
In general this means that MySQL 4.1 does not have the required stability….
Adjusting management procedures
The >anti-spam engine creates a large number of log files when storing statistics information and these files must be deleted “manually”. That is: in the local clean-up procedure that is called in the nightly batch, it is cheecked if such files exist and if so, all are deleted:
if f$search("sqlite*.*") .nes. "" then delete sqlite*.*;*
I’ve added this line some time ago but still there is quite a number of files present. No wonder: I’d better specified where to look:
if f$search("sys$manager:sqlite*.*") .nes. "" then delete sys$manager:sqlite*.*;*
👿 Duh 👿
The backup procedures have been enhanced a bit as well. Logging of the MySQL backup went to the wrong place, and only the public web is backed up – until today. I chnaged both procedures: have the logging come to the right spot, and keep only the latest web backup available, otherwise the disk wqill be crowded with unused files. They are big enough, but all tyhese backups are not needed so I can clean them up anyway – if backup succeeds.
Why did mysql-watcher not run? Well, the answer is found in the logfile:
Resubmitted 30-Nov-2007 23:32:00.00
Resubmitted 30-Nov-2007 23:47:00.00
Resubmitted 31-Nov-2007 00:02:00.00
😳 Invalid date…