MySQL processes - quickly rising

[villieb]

Morning,

I noticed this morning that Cacti wasn't logging any data in any of my graphs. I couldn't SSH to the server. I managed to get console access to the server and saw some errors (I didn't note them down )

Looking at the MySQL Process List, I am seeing right now, 112 rows of

| 455 | cacti | localhost | cacti | Sleep | 4 | | NULL |
| 459 | cacti | localhost | cacti | Sleep | 4 | | NULL |
| 461 | cacti | localhost | cacti | Sleep | 4 | | NULL |
| 463 | cacti | localhost | cacti | Sleep | 0 | | NULL |
| 470 | cacti | localhost | cacti | Sleep | 2 | | NULL |
| 471 | cacti | localhost | cacti | Sleep | 2 | | NULL |
| 473 | cacti | localhost | cacti | Sleep | 2 | | NULL |
| 475 | cacti | localhost | cacti | Sleep | 0 | | NULL |
| 478 | cacti | localhost | cacti | Sleep | 0 | | NULL |

One of the colomns, time, is as high as 1619...

Is this normal?

[villieb]

I now have 151 rows after checking the processlist...

[phalek]

What's the overall status of the server:

Diskspace ( df -h )
Processes ( ps -efwww )

Did you try a restart of mysql ?

[villieb]

Diskspace:
/dev/mapper/nhs--cacti-root 108G 2.8G 99G 3% /
udev 486M 4.0K 486M 1% /dev
tmpfs 198M 256K 198M 1% /run
none 5.0M 0 5.0M 0% /run/lock
none 495M 0 495M 0% /run/shm
/dev/sda1 228M 50M 167M 23% /boot

Processes:
Attached in process.txt

I did try restarting the sql process but the number of entries grow quickly...

[phalek]

It seems like the poller is being called either too often, or hangs.

Can you check /var/log/cacti/poller-error.log for any issues ?

[villieb]

localadmin@nhs-cacti:/var/log/cacti$ sudo tail poller-error.log
PHP Warning: mysql_pconnect(): Too many connections in /usr/share/php/adodb/drivers/adodb-mysql.inc.php on line 383
PHP Warning: mysql_pconnect(): Too many connections in /usr/share/php/adodb/drivers/adodb-mysql.inc.php on line 383
PHP Warning: mysql_pconnect(): Too many connections in /usr/share/php/adodb/drivers/adodb-mysql.inc.php on line 383
PHP Warning: mysql_pconnect(): Too many connections in /usr/share/php/adodb/drivers/adodb-mysql.inc.php on line 383
PHP Warning: mysql_pconnect(): Too many connections in /usr/share/php/adodb/drivers/adodb-mysql.inc.php on line 383
PHP Warning: mysql_pconnect(): Too many connections in /usr/share/php/adodb/drivers/adodb-mysql.inc.php on line 383
PHP Warning: mysql_pconnect(): Too many connections in /usr/share/php/adodb/drivers/adodb-mysql.inc.php on line 383
PHP Warning: mysql_pconnect(): Too many connections in /usr/share/php/adodb/drivers/adodb-mysql.inc.php on line 383
PHP Warning: mysql_pconnect(): Too many connections in /usr/share/php/adodb/drivers/adodb-mysql.inc.php on line 383
PHP Warning: mysql_pconnect(): Too many connections in /usr/share/php/adodb/drivers/adodb-mysql.inc.php on line 383

That line in the adodb-mysql.inc.php contains:

Code: Select all

if (ADODB_PHPVER >= 0x4300)
                        $this->_connectionID = mysql_pconnect($argHostname,$argUsername,$argPassword,$this->clientFlags);
                else
                        $this->_connectionID = mysql_pconnect($argHostname,$argUsername,$argPassword);

Thanks

[phalek]

Edit your /etc/my.cnf file and add a line like the following to it:

Code: Select all

max_connections = 250

then restart your mysql server. and check the logfile and processes again.

[villieb]

Thanks - I've edited the file /etc/mysql/my.cnf

So far, the processes within mysql is up at 61 rows...

I'm not seeing the errors as before in the poller log file...

Fingers crossed...

[phalek]

Ok,

If you can, maybe a system restart (or manually kill to the spine/poller processes ) to get rid of these "zombies" would be fine.

[villieb]

I've just restarted the server... output from the show processlist is empty (apart from the show processlist process)....

Lets see how this gets on now

[villieb]

Up at 33 rows when looking at the mysql processlist

[phalek]

Are there still errors in the poller-error.log showing up ?

You could as well check if there are any error messages from mysql or in your syslog.

[DrLightman]

Hello, I'm sorry for the gravedig, but I'm experiencing the very same behaviour described here and since it's 3 years old at least, I suppose it's still around.

The more the sysadmin keeps the tool open in the browser, the more mysql connections/processes spawn and don't properly close. They actually fill in the connections of the webserver and may cause other clients/processes to fail to connect in the end.



Do you think of any workaround to let it close those connections properly or maybe use one connection only.

Thank you.

[DrLightman]

Hello?

[phalek]

Do you have any plugins installed, if yes, which ones and version ?