-
Network agent overload?
Hello all!
We are proposing Pandora as the main choose for monitoring for our customers and we are going to install for some customers now.
I have encountered a problem. We are monitoring 7 network devices at the moment with the network agent.
For five of them the only check set is the “host alive check”. All works fine until I add two devices CISCO 5400 with 54 ethernet devices for each one. I have set the monitoring of the status of each port and the traffic inbound and outbound. Only for these two devices, the total amount of checks becomes 300 more or less.
The monitoring system with the network agent, after this adding, seems to be very slow and the “Out of limits” warning appears. Could it be a network agent overload? If yes, since the server is not at 100% cpu load, is it possible to start another instance of network agent to balance the operation? -
10 Replies
-
Hello all!
We are proposing Pandora as the main choose for monitoring for our customers and we are going to install for some customers now.
I have encountered a problem. We are monitoring 7 network devices at the moment with the network agent.
For five of them the only check set is the “host alive check”. All works fine until I add two devices CISCO 5400 with 54 ethernet devices for each one. I have set the monitoring of the status of each port and the traffic inbound and outbound. Only for these two devices, the total amount of checks becomes 300 more or less.
The monitoring system with the network agent, after this adding, seems to be very slow and the “Out of limits” warning appears. Could it be a network agent overload? If yes, since the server is not at 100% cpu load, is it possible to start another instance of network agent to balance the operation?Hi. You should not have a problem with this load. Pandora has been tested with more than 2500 modules in a single server. What ammount of network threads do you have assigned for the whole total of that modules ?. What version of Pandora FMS are you running and in what hardware / Linux distro ?.
Probably the most easy solution will be:
a. Increase number of network threads to …10 or 15.
b. Make a correct database purge each night or almost one per week. -
-
-
-
-
-
Next Friday I will try nil’s suggestion.
Meanwhile… Manu, do you think that 1.3.1 improves or adjusts the problem I have encountered?
There is not any difference in performance between 1.3.1 and 1.3, so thread and timeouts should fix your problem. 1.3.1 comes with some hacks and additional parameters to fine-tune network monitoring, but core network procesor is the same.
But 1.3.1 is now much better (and stable) than official 1.3 version, so try it 😉
-
There is not any difference in performance between 1.3.1 and 1.3, so thread and timeouts should fix your problem.
Ok, database purged. Network Threads modified from 5 to 15 and then to 30.
No differencies shown yet.When you talk about timeouts, do you talk about the Network_timeout in the pandora_server.conf or the timeouts in the settings of each agent?
-
-
