To prevent query sampling from using an unreasonable amount of disk space when a user forgets to turn off query sampling after they finish analyzing a shard key or when the sample rate is low but the queries are just so large in size, we should:
- The config.sampledQueries and config.sampledQueriesDiff should exist as capped collections, with a size cap of 10GB.
- Set the upper bound for the sampleRate to 1000.
EDIT: Please refer to the first comment in this ticket. This ticket only enforces a low upper bound for the sample rate.
- split to
-
SERVER-76477 Cap the size of the collections storing sampled queries and diffs
-
- Backlog
-