[SERVER-38441] In-Memory pins all dirty data in cache because of read concern majority Created: 06/Dec/18  Updated: 27/Oct/23  Resolved: 18/Dec/18

Status: Closed
Project: Core Server
Component/s: Replication, Storage
Affects Version/s: 3.6.9, 4.0.4
Fix Version/s: None

Type: Bug Priority: Major - P3
Reporter: Danny Hatcher (Inactive) Assignee: Tess Avitabile (Inactive)
Resolution: Works as Designed Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Attachments: PNG File Screen Shot 2018-12-06 at 3.39.57 PM.png     PNG File Screen Shot 2018-12-06 at 3.41.05 PM.png    
Issue Links:
Related
related to SERVER-38685 Startup warning if In-Memory SE is us... Closed
Operating System: ALL
Steps To Reproduce:

1. Launch a replica set using the In-Memory storage engine without explicitly disabling read concern majority.
2. Perform a significant amount of inserts.
3. View that dirty pages are not evicted.

Sprint: Repl 2019-01-14
Participants:
Case:

 Description   

By default, the In-Memory storage engine runs into read concern majority cache problems. If read concern majority is not explicitly disabled at startup, all dirty data remains pinned in cache because In-Memory never commits the data as the normal WiredTiger engine does.

Default settings:

--enableMajorityReadConcern false:



 Comments   
Comment by Danny Hatcher (Inactive) [ 18/Dec/18 ]

This cache pinning is the result of writeConcernMajorityJournalDefault being set to true when using the default replica set configuration. Per the documentation, this config value should be set to false if any member of the replica set is using the In-Memory storage engine.

Opened SERVER-38685 to discuss adding a startup warning.

Generated at Thu Feb 08 04:48:58 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.