[SERVER-32442] Don’t concatenate unique fields of LinkStoreMongoDb documents to use for _id Created: 21/Dec/17  Updated: 06/Dec/22

Status: Backlog
Project: Core Server
Component/s: Replication
Affects Version/s: None
Fix Version/s: None

Type: Task Priority: Major - P3
Reporter: William Schultz (Inactive) Assignee: Backlog - Replication Team
Resolution: Unresolved Votes: 0
Labels: Linkbench
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Related
related to SERVER-32320 Push applyOps oplog entry generation ... Closed
related to SERVER-32342 Make doTxn rewrite non-idempotent wri... Closed
Assigned Teams:
Replication
Sprint: Performance 2018-05-07
Participants:

 Description   

For the MongoDB linkbench POC, the applyOps command was used to emulate transactions, but, for document updates within an applyOps command, there is no support for querying a document by anything other than it's _id field, even if the query would uniquely identify a document. This seems like a restriction that shouldn't be present for multi-statement transactions. Once transactional updates support querying unique documents by non _id fields, we should change the schema of the Link documents and Count documents, so they don't have to awkwardly concatenate their unique fields for their _id. Removing the concatenated _id should make the benchmark more analogous to the MySQL schema used for its benchmark.



 Comments   
Comment by Henrik Ingo (Inactive) [ 07/Jan/20 ]

Hi

This was done when we updated the code to use real transactions. This ticket can be closed.

See https://github.com/10gen/linkbench/blob/master/README.md#running-a-benchmark-with-mongodb

Comment by Ian Whalen (Inactive) [ 21/May/18 ]

tess.avitabile this doesn't need to be done as part of the transaction perf support project, so assigning back to repl for triage.

Generated at Thu Feb 08 04:30:15 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.