Database disconnects zabbix.

[mrmurdock]

Zabbix_1.6.6
Hi, I have increased the number of connections in my sql to 2048, but for some reason I still get the below error message...I am only running at a max 102 zabbix_server processes and could increase up to 200 (im assuming 1 zabbix_server process per host, whether zabbix_agentd or icmpping).


mysql_pconnect() [<a href='function.mysql-pconnect'>function.mysql-pconnect</a>]: Too many connections[/home/zabbix/public_html/include/db.inc.php:56]

[richlv]

wait, what do you mean by 102 zabbix_server processes ? that sounds like an insanely high amount...

[mrmurdock]

I just updated to zabbix 1.6.7 - still the same thing happens

When I do a ps ax | grep zabbix_server | wc -l I get back 94-102.

Each host (zabbix_agentd.conf) is pre-configured for 3-5 agentd's (not really sure why one would need any more than 1 zabbix_agentd process running per host).

Not sure what connections are connecting into the DB (mysql 5.0 RHEL5).

From zabbix Monitoring --> Dashboard
I have 141 hosts (81 monitored, 17 not monitored (ping'ed ??),43 templates).
I have 4642 items (3520 monitored, 189 not, 933 unsupported).
I have 1605 Triggers (965 enabled, 640disabled) (48true,153unkown,764False)
1 User logged in.

here is the output of ps on mysqld
15685 ? Sl 24:57 /usr/libexec/mysqld --defaults-file=/etc/my.cnf --basedir=/usr --datadir=/var/lib/mysql --user=mysql --pid-file=/var/run/mysqld/mysqld.pid --skip-locking --open-files-limit=2048 --socket=/var/lib/mysql/mysql.sock

[richlv]

When I do a ps ax | grep zabbix_server | wc -l I get back 94-102.

ouch. how many pollers have you configured for your zabbix server ?

[data7]

&quot;Disconnection&quot; problem

Hello, I had a similar problem with 1.6.6 and updating to 1.6.7 didn't solve it at all.

I don't know if its a known bug, but from 1.6.5 onwards I have been receiving false e-mail notifications containing "Zabbix database is down" alerts. There was no database disconnection ou interruption on the Zabbix daemon at all.

Odd fact: These messages come from most of my child nodes at an EXACT time! Some of them are not on the distributed setup though.

This is bad since an alert like that can wake up lots of people (including me).

Anyone else got this issue? Haven't checked this on the bug tracking system.

I set these timeout parameters to the maximum:
Timeout=30
TrapperTimeout=300

And both pollers and trappers are below 20. The servers aren't overloaded as well.

[trikke]

Hi,

Are u doing a DB-Backup or whatever on your DB? ( That's what I have! )
Is there something locking your DB? ( Housekeeping, perhaps?)

greets Patrick

[data7]

Hi,

Are u doing a DB-Backup or whatever on your DB? ( That's what I have! )
Is there something locking your DB? ( Housekeeping, perhaps?)

greets Patrick

Nope...But now that you mentioned, I noticed that the time the notification is sent matches with the start of the Housekeeping process on the master node...which is even stranger since I got notifications from (future) child nodes that haven't been added to him yet!

I have no idea of what's generating these events...Hope it's not related to security, since most replication data flow through the internet without encryption...

I am getting these events from servers using 1.6.5, 1.6.6 and 1.6.7.