I have around 300 Mikrotik routers currently being monitored in a test environment. 100 other types of devices. I was noticing that the queue was falling behind so I looked at the templates to see how often it polls certain metrics. I noticed that the Mikrotik templates have SNMP walk set to every minute for interfaces, CPU, temperature, and something else.
I decided to make a macro for POLLING_INTERVAL then change them from 1m to 5m just to see how it handled it. This really seemed to drop the query load a bunch.
Then I checked the Juniper profile and the Brocade profile to see what their walk time was set for and how they were handling it. I found they were not walking interfaces at all in their "items" section. Instead they have a Discovery rule that performs the walk every hour. The interfaces are added as item prototypes with triggers for high bandwidth usage and other things. The individual metrics are set to poll every 3 minutes in the item prototype.
So I was wondering why Mikrotik wasn't like this. Is there something about LLD that Mikrotik didn't support so it had to be done manually? Or are the mikrotik templates old and haven't been updated yet?
I checked the latest git version of the Mikrotik templates and, although I didn't load them into a Zabbix instance, they appear to be working the same way.
I also didn't bother to load specific templates for the specific Mikrotik versions. I did look at their YAML to see that they seem to work the same way. If there is a chance that will make things better I can update our discovery rule to try to load the specific templates.
Has anyone else noticed this and how are you handling Mikrotik polling?
I decided to make a macro for POLLING_INTERVAL then change them from 1m to 5m just to see how it handled it. This really seemed to drop the query load a bunch.
Then I checked the Juniper profile and the Brocade profile to see what their walk time was set for and how they were handling it. I found they were not walking interfaces at all in their "items" section. Instead they have a Discovery rule that performs the walk every hour. The interfaces are added as item prototypes with triggers for high bandwidth usage and other things. The individual metrics are set to poll every 3 minutes in the item prototype.
So I was wondering why Mikrotik wasn't like this. Is there something about LLD that Mikrotik didn't support so it had to be done manually? Or are the mikrotik templates old and haven't been updated yet?
I checked the latest git version of the Mikrotik templates and, although I didn't load them into a Zabbix instance, they appear to be working the same way.
I also didn't bother to load specific templates for the specific Mikrotik versions. I did look at their YAML to see that they seem to work the same way. If there is a chance that will make things better I can update our discovery rule to try to load the specific templates.
Has anyone else noticed this and how are you handling Mikrotik polling?
Comment