diff options
| author | Sebastian Mayr <smayr@atlassian.com> | 2020-10-14 18:10:01 -0400 |
|---|---|---|
| committer | Evergreen Agent <no-reply@evergreen.mongodb.com> | 2020-10-15 16:31:09 +0000 |
| commit | 14bfbd8833706912617c0c904cce9847fbbbd0f1 (patch) | |
| tree | e8876ba38ae5c0d79f458f100a4e684e68207b3e /jstests/sharding/after_cluster_time.js | |
| parent | 4772fe5bd1a4e66c5cc8177cdb5956e05b07b176 (diff) | |
| download | mongo-14bfbd8833706912617c0c904cce9847fbbbd0f1.tar.gz | |
SERVER-39191 Optimize count queries on sharded collections
When counts were made orphan-aware the addition of the SHARD_FILTER
stage started preventing the COUNT_SCAN optimization on all sharded
collections, leading to significant performance loss on some workloads.
However, there can not be any orphans if the shard key is fully
specified: A mongod will never have orphans for chunk ranges
it owns.
This patch changes the query system to omit the SHARDING_FILTER stage in cases
where the query specifies the full shard key, regaining lost performance.
Closes #1369.
Signed-off-by: Ian Boros <ian.boros@mongodb.com>
Diffstat (limited to 'jstests/sharding/after_cluster_time.js')
0 files changed, 0 insertions, 0 deletions