Unfortunately, the agent baking setting does not noticably change the behavior for us. We’re still talking about 40 seconds or more for every activation, no matter what changes are made and how many hosts / services are affected by the changes.
I’m wondering how we would go about zooming into the issue - backup the entire instance, restore on another system, move all custom content (custom checks, web\wato, …) out of ~/local/ and cmk -R to get a “vanilla” Checkmk install?
Why are we talking about cmk -R in this thread btw, does a GUI “activate” not trigger a “cmk -O” by default, reloading instead of restarting? cmk -O = 26 seconds, cmk -R = 40 seconds (as it entails shutting down all the processes and spawning an entirely new master process that needs to process all the filesystem config and state files anew), so it’s a Restart apparently every time we do a GUI Activation, even for typo fixes in Host aliases… Reload duration would be more palatable, but is still 2x-3x as slow as with 2.x and previous versions