[SERVER-15950] Reduce contention on groupCommitMutex in MMAP V1 Created: 04/Nov/14  Updated: 19/Nov/14  Resolved: 06/Nov/14

Status: Closed
Project: Core Server
Component/s: Concurrency
Affects Version/s: 2.7.8
Fix Version/s: 2.8.0-rc0

Type: Bug Priority: Major - P3
Reporter: Davide Italiano Assignee: Mathias Stearn
Resolution: Done Votes: 0
Labels: 28qa
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Attachments: PNG File vtune_profiler_groupcommit.png    
Issue Links:
Related
Backwards Compatibility: Fully Compatible
Operating System: ALL
Steps To Reproduce:

Run sysbench (https://github.com/tmcallaghan/sysbench-mongodb)

Participants:

 Description   

On insert-only workload (e.g. sysbench first phase), groupCommitMutex shows up as the most congested lock. Attached the Vtune lock profiler output as an evidence



 Comments   
Comment by Githook User [ 06/Nov/14 ]

Author:

{u'username': u'RedBeard0531', u'name': u'Mathias Stearn', u'email': u'mathias@10gen.com'}

Message: SERVER-15950 Optimize handing of writingPtr

Comment by Kaloian Manassiev [ 05/Nov/14 ]

Note that this is specific to MMAP V1, in case you were trying with other storage engines. Disabling journaling should help.

Comment by Tim Callaghan [ 05/Nov/14 ]

Is this gating the performance improvement I'd expect to see with collection level locking? Is disabling journaling a work-around for now?

Generated at Thu Feb 08 03:39:30 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.