-
Type: Improvement
-
Resolution: Done
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: Storage
-
None
-
Fully Compatible
-
Storage 2017-04-17, Storage 2017-05-08
-
(copied to CRM)
There is a point where there is a significant performance decrease with WiredTiger storage engine if too much content is required to remain in cache to service old operations. The reason for the performance difference is that WiredTiger starts paging content out from cache to disk in temporary files, when old updates may be required for old readers, but are no longer maintained in the database file.
The read-concern-majority functionality can cause a lot of updates to be pinned in cache, especially if there is significant lag in a replica set using read-concern majority.
The goal of this ticket is to construct a workload that forces an excessive amount of history to be generated and to assess the behavior of the system under such circumstances.
- depends on
-
WT-3296 LAS table fixes/improvements
- Closed
- related to
-
SERVER-31099 Automate testing when oldest_timestamp stalls
- Closed