On 2022-10-19, Alec <olp...@yahoo.ca> wrote: > Hi, > Thank you for a quick reply! That is a bit sad that we only have coldStart, > but I udnerstand. > It would be great to have traps for CPU/temperature threshold breaches for > general stuff and addition (especially) as well as deletion of IP addresses > from a given table.
PF tables are polled by snmpd when the relevant MIBs are requested, there's no 'push notification' from the kernel about changes to these. While not impossible this would require fairly wide ranging changes to implement in a sane way. (snmpd can't really just keep fetching the tables periodically itself and compare them to find changes...) Temperature alerts are indeed something that would be useful, though there isn't a notion of "what temperature is too high/low" in the kernel framework (other than "does the sensor itself report 'critical'" which IME is not actually very useful). This would probably best be done in conjunction with sensorsd and its configuration language (perhaps even moving the whole sensors mib there from snmpd-metrics, though that might be overkill). However I think you could /probably/ already implement this yourself by calling snmp(1) 'trap' from sensorsd if you wanted this now. > For future references, are there any specific traps you would like to > see implemented? Not a high priority, but the most useful for me would be linkDown, linkUp (.1.3.6.1.6.3.1.1.5.3 and .4), and BGP neighbour state (bgpEstablished, bgpBackwardTransition from BGP4-MIB). The latter being more useful (as the link states can be picked up easily enough by polling).
