The maximum load that can support a Pandora FMS environment will depend on the following characteristics that are presented in the environment:
- Capacity of the hardware and software installed. When more CPU capacity, RAM and speed of reading and writing to disk has the server hosted more modules / second can process it. Starting from an environment of 2 CPU / 2 GB RAM and 7200 RPM disk that can support about 100 Agents and 1000 modules without problems, to servers with much higher capacity that can get to process more than 10000 Agents without problems.
- Monitoring interval. The parameter that will measure the performance that can support the environment is the number of modules / second that you need to support. It is not the same to have 1000 modules at 1 minute update interval, than 1000 modules at 10 minutes interval. The first one will need to process 16 modules per second and the second one 1.6 modules per second.
- Network latency. In environments with higher network latency the number of agents and modules that can be monitored is lower than the number of elements that will be monitored in an environment with low latencies. Starting from the premise of number of modules/second, if for each check we have to spend 2-3 seconds instead of 0.1 second, the maximum number of network checks that will be done will be lower and this will limit the capacity of it. This problem can be solved by installing more servers.
- Architecture with several servers: Spreading the load of Apache, MySQL and Pandora FMS processes in several servers will help us to get a higher rate of modules per second and our environment will be able to support a higher load.
- Server optimization. One of the most important points is to have a correct optimization of PHP, MySQL and Pandora FMS server according to the environment to be monitored and the available resources of the server. It is not worth having a very powerful server if we use the basic configuration of MySQL, PHP or the server itself and we do not take advantage of the available resources.
- Data history. The more agile the main database is and the less data it has, the faster the monitoring processing will be. Therefore, the use of historical databases is recommended for environments with large data storage needs
- Considering all these metrics it is essential that the total server load calculated with the maximum module processing rate per second does not exceed 60-75% of the maximum supported load. In the case of not complying with this premise we may find ourselves in moments when there is an overload in the environment when processing an important report and we have delays in monitoring.
For example, if an environment has a rate of 20 modules/second of processing and we have an interval of 5 minutes configured in all the modules, the maximum number of modules that will be processed is 20*300=6000 modules. Therefore, it is recommended that if we set the threshold to 60%, the maximum number of modules configured in that environment will be 6000*0.6=3600 modules.
To calculate the module rate that supports an environment we have the script
/usr/share/pandora_server/util/pandora_dbstress.pl /etc/pandora/pandora_server.conf
installed with the Pandora FMS server itself that gives us the maximum processing rate:
-> Current rate: 1989.03 modules/sec.