Description Backlog allowed before canceling the queries with the highest cost

In combination with LIVEQUERY_STALE_CANCEL_COST_PERCENTAGE, controls how stale queries are managed:

  • The "Stale" monitoring is concerned with queries that have not been polled recently. Those get cancelled after a while if their cost sum is sufficiently high. This is done on the coordinator side. The internal task is LivequeryMonitor.

  • The "non stale" monitoring is concerned with the local query execution in a worker node where it inspects the cost of all queries running and starts cancelling some if the ingest delay increases above the threshold set by LIVEQUERY_CANCEL_TRIGGER_DELAY_MS The internal task is LocalLivequeryMonitor. Note how this one may drop queries even if they are still being used or viewed.