[SERVER-17118] moveChunk is slow with wired tiger Created: 29/Jan/15  Updated: 09/Jul/16  Resolved: 02/Feb/15

Status: Closed
Project: Core Server
Component/s: Sharding, WiredTiger
Affects Version/s: 3.0.0-rc7
Fix Version/s: None

Type: Improvement Priority: Major - P3
Reporter: Siyuan Zhou Assignee: Andy Schwerin
Resolution: Duplicate Votes: 0
Labels: 28qa
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Attachments: PDF File move_from.nodelete.pdf     PDF File move_to.nodelete.pdf    
Issue Links:
Depends
Related
related to SERVER-17151 Avoid fsync command in ConfigCoordinator Closed
Backwards Compatibility: Fully Compatible
Participants:

 Description   

Comparing to 2.6 and MMAPv1, moveChunk is slow with wired tiger.

Version shard01 => shard02 _waitForDelete: false
2.6 7017 ms
MMAPv1 8150 ms
WiredTiger 21762 ms


 Comments   
Comment by Siyuan Zhou [ 02/Feb/15 ]

Further investigation shows that it's ConfigCoordinator who sends "fsync" commands to config servers, not SyncClusterConnection. SyncClusterConnection uses { "getlasterror": 1, "fsync" : 1 } instead, which doesn't flush files if journaling is enabled. ConfigCoordinator is used by both mongos and mongod for writing to config servers.

SERVER-17151 has been filed to track this bug. Perhaps we can close this ticket as a dup since the most important reason is SERVER-17151.

Generated at Thu Feb 08 03:43:21 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.