I am troubleshooting an issue with slow queries in my Prometheus-compatible datasource. When I identify an offender, I can normally track down the explore URL or Grafana dashboard from which the query came, and Grafana access logs tell me who has accessed those URLs.
Now, however, the MCP mixes all the requests together and I can't track where they originated from. The debug log is too verbose to be useful for this purpose.
Therefore, add a slow query log. An optional configuration parameter can be set so that queries/requests (of any kind, not just PromQL queries) can be logged as info-level or warning-level log events if they take longer than the specified amount of time.
I am troubleshooting an issue with slow queries in my Prometheus-compatible datasource. When I identify an offender, I can normally track down the explore URL or Grafana dashboard from which the query came, and Grafana access logs tell me who has accessed those URLs.
Now, however, the MCP mixes all the requests together and I can't track where they originated from. The debug log is too verbose to be useful for this purpose.
Therefore, add a slow query log. An optional configuration parameter can be set so that queries/requests (of any kind, not just PromQL queries) can be logged as info-level or warning-level log events if they take longer than the specified amount of time.