I have a environment with a central server (Web UI, DB, rrd,…) and 2 Pollers
whats is the best way to use the pollers? Each one in a poller group or both in the same poller group?
How are the devices distributed over the poller server, wenn they are in the same group?
What is going on, when both poller servers are in one group and one pollerserver is defect? Are then all devices polled by the remaining server in this group even if the remaining server has not enough power to poll all devices?
Depending on how many workers each poller has, if they are even then it’s roughly a 50/50 split.
Yes they are all polled by the other poller (if it has capacity).
A lot of time people put pollers in different groups for isolation or seperation reasons (maybe geographically). However you can do both, you can have multiple pollers in multiple groups.
For example let’s say you have the East and West coast of the states.
You could put one poller on the west and one on the east in the same group. However the east poller will have higher latency to the west devices so may not be ideal but if one poller fails, you still have one remaining.
The other options is two pollers on the west coast and two on the east coast. You then have two groups, East and West with the respective pollers in. Latency isn’t an issue for the pollers and you have resilience.
One caveat to all of this, remember you will have rrd and mysql latency back to the central DB / RRD server that you need to account for.