[SERVER-36125] Race in moveChunk and createIndex can cause a shard to have missing indexes Created: 13/Jul/18  Updated: 06/Dec/22  Resolved: 13/Jul/18

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

Type: Bug Priority: Major - P3
Reporter: Randolph Tan Assignee: [DO NOT USE] Backlog - Sharding Team
Resolution: Duplicate Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Duplicate
duplicates SERVER-31715 createIndexes (and dropIndexes) may n... Closed
Assigned Teams:
Sharding
Operating System: ALL
Participants:

 Description   

... until next migration.

1. moveChunk gets called, destination shard calls listIndexes.
2. createIndex with text index was sent to recipient shard.
3. destination shard got the list of indexes, but doesn't include the in progress text index.
4. text index finished in recipient.
5. moveChunk completes
6. text query get targeted to the destination shard - error: text index required



 Comments   
Comment by Kaloian Manassiev [ 13/Jul/18 ]

esha.maharishi, didn't you discover something like that in 3.6 and filed a ticket for it? This is a known deficiency of the migration destination manager that can't be fixed unless we do networking operations under a lock :/

Generated at Thu Feb 08 04:42:06 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.