Werk #22178: Kubernetes: Fix missing CPU usage when the cluster collector returns unchanged data
| Component | Checks & agents | ||||
| Title | Kubernetes: Fix missing CPU usage when the cluster collector returns unchanged data | ||||
| Date | Aug 17, 2026 | ||||
| Level | Trivial Change | ||||
| Class | Bug Fix | ||||
| Compatibility | Compatible - no manual interaction needed | ||||
| Checkmk versions & editions |
|
CPU usage is a rate that must be calculated from two samples taken at different points in time. These samples are sent to the cluster collector from the node collectors, where they are cached until Checkmk queries for the data.
In the case where Checkmk queried the cluster collector twice before the node
collector had a chance to update the data stored in the cluster collector,
both samples were identical and no rate could be computed. This scenario would
often cause the CPU resources service to stop updating and go stale. The
corresponding resource quota service, Resource quota cpu resources, could
stop being executed altogether, because for a resource quota that does not
define hard CPU values, this performance data is the service's only data
source. In that case, the Check_MK service would report
"Missing monitoring data for plugins, kube_resource_quota_cpu".
The special agent now stores the last rate it computed for each container and reuses it for up to 5 minutes while the cluster collector keeps returning unchanged samples. The sections therefore remain present and the services keep being executed.
Note that during such a period, the previously computed value is displayed instead of a gap and levels are evaluated against that previous value. If no new data arrives from the cluster collector for more than 5 minutes, the section is omitted as before.
No user action is required to make use of this fix.