Details
Description
MongoDB serializes the creation of new WiredTiger transactions for local secondary reads (i.e. lastApplied reads) and readConcern: majority reads.
WT-6709 recently addressed performance problems in the WiredTiger timestamp queues that should make these mutexes unnecessary. Investigate removing both of these mutexes to potentially support higher read throughput.
SeeĀ SERVER-51041 for more details.
Attachments
Issue Links
- causes
-
SERVER-63244 Majority reads must round their read timestamp up to the oldest timestamp
-
- Closed
-
- is depended on by
-
SERVER-55024 Complete TODO listed in WT-6709
-
- Closed
-
- is related to
-
WT-9074 Reconsider spinlocks in transaction ID allocation path
-
- Open
-
- related to
-
SERVER-51041 Throttle starting transactions for secondary reads
-
- Closed
-
-
WT-6709 Remove timestamp queues that used to store read/durable timestamps
-
- Closed
-