I manage a monitoring setup running Icinga2 version r2.6.3-1 (from the 2.6.3-1~debmon8+1 debian package) in a 2 node master-client setup with 153 hosts and 1456 services on a postgresql bdr database.
After the icinga2 process is started or reloaded, the cpu usage is normal (15%). After an arbitrary amount of time (between ~10 min and ~6 hours), the cpu load of the icinga2 process jumps to 100% and the icinga2 query queue starts rising, complaining about "empty in infinite time, your database isn't able to keep up". It will remain like this until the next reload, at which time the old process will stop monitoring and quickly empty its query queue into the database and a new process is forked which starts monitoring like it should. For as far as I've been able to see no data gets lost. Also monitoring gets done normally both in the normal state and in the high cpu state.
A dirty work around for this issue could be accomplished by monitoring the icinga2 log and letting icinga2 restart itself whenever it finds a complaint about the time it expects to empty its queue. A real fix is preferred :-).
I've not been able to find a recent post with similar issues, though it could be (related to) this bug: https://github.com/Icinga/icinga2/issues/4133
What action/info would you need on my end to help fix this?