SNMPAgent: Unable to find a poller

[TheWitness]

Cool.

[cedricpah]

Hi, I got the same with log message 40000 max ect...
Fin in log cacti_stderr.log that 'FATAL: Connection Failed: Error '1040', Message:"Trop de connexions"
So I put in my mariadb config file (mine is in /etc/mysql/mariadb.conf.d/50-server.cnf) a line with max_connections=500 (default is 100) and this now works without error message!

[TheWitness]

Better to open another thread for this. Using old threads is bad karma. This is a common problem. Often times it happens when users are new to Cacti and try to use spine and do do the math on the number of connections that are required for Cacti's use. There is a setting max_connections in MySQL/MariaDB that is very important. If using spine, Cacti needs a lot of them. Remember, if you are only monitoring a few thousand hosts, you don't need too many processes, you just have to balance the number of threads + processes to be somewhere between 2-4 times the number of CPU cores. If the system is real big, then you have to to timing studies based upon response times of your agents, and the WAN latency between your Cacti server and it's devices and find the sweet spot, but under no condition should ANYONE use something like:

10 processes
20 threads per process.

That requires >= 200 threads, and most versions of MariaDB/MySQL, though they can handle this load, don't do it well. When you feel you've gotten there, you need to distribute your pollers. I suspect you are nowhere near that.

https://stackoverflow.com/questions/143 ... connection