Werk #22465: Rate-based services of push hosts no longer go stale
| Component | Core & setup | ||
| Title | Rate-based services of push hosts no longer go stale | ||
| Date | Oct 28, 2026 | ||
| Level | Trivial Change | ||
| Class | Bug Fix | ||
| Compatibility | Compatible - no manual interaction needed | ||
| Checkmk versions & editions |
|
Previously, some services of hosts using the push agent became stale although every push arrived on time. This affected services based on rates, such as CPU utilization, Disk IO SUMMARY and Kernel Performance.
The agent pushes its data every 60 seconds and the Check_MK service of the host also runs every 60 seconds. When a push arrived just after a check, the next check read the same agent data again. Rate-based checks cannot compute a rate from unchanged data, so they produced no result in that cycle. Their last result was then up to 120 seconds old, which is above the default staleness threshold of 1.5 check intervals.
The data of push agents is now reported to the core as cached since its arrival at the site, for the push interval of 60 seconds. This is similar to how piggyback data is handled. The staleness of these services is now computed against the check interval plus the push interval, so a single skipped result no longer makes them stale. The fix requires the Checkmk Micro Core (CMC).
If the pushes stop, the Check_MK service still goes CRIT as before. The
other services now go stale 180 seconds after the last push arrived, instead
of after 90 to 180 seconds.
As a side effect, the services of push hosts now show the age of the agent data. Instead of the button to reschedule the Check_MK service, they show an icon saying that they are based on cached agent data and cannot be rescheduled.