[SERVER-20201] Performance drop over time when running with enableReplSnapshotThread=true under insert only workload Created: 30/Aug/15  Updated: 06/Dec/22  Resolved: 06/Sep/16

Status: Closed
Project: Core Server
Component/s: WiredTiger
Affects Version/s: None
Fix Version/s: None

Type: Bug Priority: Major - P3
Reporter: Eitan Klein Assignee: Backlog - Storage Execution Team
Resolution: Duplicate Votes: 0
Labels: 32qa
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Attachments: HTML File 318.html     PNG File Compare.png     PNG File profiler.png     Text File rs3primary.txt    
Issue Links:
Duplicate
duplicates SERVER-20174 WT snapshot threads significantly imp... Closed
Related
Assigned Teams:
Storage Execution
Backwards Compatibility: Fully Compatible
Steps To Reproduce:

Replica set
Workload - Hammer.mongo, insert only profile w/ write majority

%MongoD%\mongod --wiredTigerCacheSizeGB 3 --replSet EitanRs3b --dbpath f:\data\db1 --logpath h:\data\rs3primary.txt --storageEngine wiredTiger --port 5002 --setParameter enableReplSnapshotThread=true

Participants:

 Description   

Environment:

2 member replica set (Windows)
db version v3.1.8-pre-
git version: 9b054f00326b23a87ba0571cbd8997b91fe5ce16

Storage size: 13 GB
Workload:

  1. Hammer.mongo, insert only profile with write concern majority
  2. 16 threads
  3. Execution time 12 hours

Problem:

  • The same workload w/ enableReplSnapshotThread=false produce significant better throughput - See compare.png

The profiler indicates high CPU as results of __wt_tree_walk function that accountable for 80.83% of the CPU time - See profiler.png


Generated at Thu Feb 08 03:53:29 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.