[SERVER-44913] Index operations should only abort migrations of first chunk to a shard Created: 02/Dec/19  Updated: 06/Dec/22  Resolved: 14/Feb/20

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

Type: Task Priority: Major - P3
Reporter: Jack Mulrow Assignee: [DO NOT USE] Backlog - Sharding Team
Resolution: Won't Fix Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Assigned Teams:
Sharding
Participants:

 Description   

From the design for Each Shard has Correct Indexes:

Serializing entire migrations and index operations is only necessary when the recipient is receiving its first chunk because the shard versioning protocol will guarantee all shards with chunks will be targeted by index operations. So as an optimization, only the migration of the first chunk to a shard needs to be serialized with index operations (beyond existing serialization around entering the migration critical section).



 Comments   
Comment by Esha Maharishi (Inactive) [ 14/Feb/20 ]

Closing because this was not a requirement for correctness and was an optimization that was difficult to do. We can reopen if needed.

Comment by Jack Mulrow [ 22/Jan/20 ]

This isn't necessary for correctness, so moving out of PM-1361. It might still be useful to do this if users are seeing many migrations abort because of concurrent index operations, so moving to the general sharding backlog for triage.

Generated at Thu Feb 08 05:07:21 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.