Werk #22091: bi: time out slow aggregation state requests
| Component | BI | ||||
| Title | bi: time out slow aggregation state requests | ||||
| Date | Aug 4, 2026 | ||||
| Level | Prominent Change | ||||
| Class | Bug Fix | ||||
| Compatibility | Compatible - no manual interaction needed | ||||
| Checkmk versions & editions |
|
When a Business Intelligence special agent was configured, the REST API requests it made had no timeout. A slow aggregation compilation or state fetch occupied an Apache worker indefinitely: the core stopped waiting for the special agent once its check timeout was reached, but the site kept working on the abandoned request. With every check interval adding another one, the workers were used up until the site stopped answering at all.
Computing the aggregation state is now bounded at 90 seconds. The endpoints
cmk/bi_aggregation_state_get and cmk/bi_aggregation_state_post answer with
504 Gateway Timeout once that bound is reached, which frees the worker. Where
the check timeout allows the special agent to wait that long, it reports the
error in its service output instead of hanging.
Note that aggregations which need more than 90 seconds to compute now report this error rather than a state. This was previously only reachable by raising the check timeout of the special agent's service well above its default of 60 seconds.