[SERVER-39458] Add continuous draining on secondary's index build thread while it awaits a commitIndexBuild oplog entry Created: 08/Feb/19  Updated: 29/Oct/23  Resolved: 03/Apr/20

Status: Closed
Project: Core Server
Component/s: Storage
Affects Version/s: None
Fix Version/s: 4.4.0-rc1, 4.7.0

Type: Task Priority: Major - P3
Reporter: Dianna Hohensee (Inactive) Assignee: Louis Williams
Resolution: Fixed Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Backports
Depends
depends on SERVER-39239 Two-phase index builds on secondaries... Closed
Related
related to SERVER-39457 Secondaries should check index build ... Closed
related to SERVER-48419 Extend rollback to recover resumable ... Closed
related to SERVER-51008 adjust rollback to resume index build... Closed
Backwards Compatibility: Fully Compatible
Backport Requested:
v4.4
Sprint: Execution Team 2020-04-06, Execution Team 2020-04-20
Participants:
Story Points: 8

 Description   

During secondary oplog application, the IndexBuildsCoordinator should periodically drain the side tables while awaiting the commitIndexBuild or abortIndexBuild oplog entries from the primary.



 Comments   
Comment by Githook User [ 10/Apr/20 ]

Author:

{'name': 'Louis Williams', 'email': 'louis.williams@mongodb.com', 'username': 'louiswilliams'}

Message: SERVER-39458 Continuously drain side writes while waiting for next index build action

(cherry picked from commit ca49965d8d68cc853e466ba741df08bb248d46f6)
Branch: v4.4
https://github.com/mongodb/mongo/commit/98be6002347d03644c7a68259ac5b0930edc6ba7

Comment by Githook User [ 03/Apr/20 ]

Author:

{'name': 'Louis Williams', 'email': 'louis.williams@mongodb.com', 'username': 'louiswilliams'}

Message: SERVER-39458 Continuously drain side writes while waiting for next index build action
Branch: master
https://github.com/mongodb/mongo/commit/ca49965d8d68cc853e466ba741df08bb248d46f6

Comment by Benety Goh [ 23/Sep/19 ]

Not required at this moment. Putting back on backlog.

Comment by Eric Milkie [ 08/Feb/19 ]

At the end of every batch would be a good cadence, as it would guarantee if there were new writes to drain, we would always drain them in a timely manner. If there are no batches happening, we can just block instead of spinning.

Comment by Dianna Hohensee (Inactive) [ 08/Feb/19 ]

milkie, you had some ideas a while back about what would be a good cadence at which to periodically re-drain the side table. Could you try to recall what they were and add those details here?

Generated at Thu Feb 08 04:52:07 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.