Big requests
This mechanism allows to cap the number of simultaneous high consumption requests, consuming much fewer resources and considerably increasing the cluster’s scalability and availability. Demands with high consumption are queued until a big request slot frees up.
Configuration
This feature is triggered when a request consumes more than query.high-cost.memory.mb
memory. You can find a comprehensive list of configuration parameters using this feature in galactica.conf.
Description
Each time a request reaches a memory consumption of query.high-cost.memory.mb
, this request is identified as a "big request".
The cluster allows no more than queries.high-cost
simultaneously running requests. Pending "big requests" will be frozen, until a big request slot is available.
The cluster allows no more than queries.high-cost.frozen
frozen requests (pending for an available slot). When the system reaches maximum pending big requests it will display this error message: System overheat, too many high-cost requests
Please note that the parameter queries
also allow you to limit the total number of queries running in parallel (small or big requests).