Currently, range deletion gets stuck searching for an index to complete an indexScan for deletion. In the case of a hashed shard key index, we will allow skipping range deletion for the collection. This may lead to orphan documents remaining on the shards, but we will solve this with documentation by recommending that a user drain orphan documents before deleting their index.
- causes
-
SERVER-99623 High CPU usage in range deleter service
-
- Backlog
-
- is related to
-
-
- Closed
-
-
SERVER-78863 Audit places in the codebase that rely on the hashed index to ensure correct behavior.
-
- Closed
-