[SERVER-33535] Add microbenchmark tests for mixed entries (old and new) in new unique index Created: 28/Feb/18  Updated: 01/Aug/18  Resolved: 01/Aug/18

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

Type: Improvement Priority: Major - P3
Reporter: Sulabh Mahajan Assignee: Sulabh Mahajan
Resolution: Won't Fix Votes: 0
Labels: nonnyc, storage-engines
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Depends
depends on SERVER-34489 Enable new format Unique Index via FCV Closed
Related
is related to SERVER-32755 Expand Evergreen microbenchmark cover... Closed
Sprint: Storage Non-NYC 2018-04-09, Storage Engines 2018-07-30, Storage Engines 2018-08-13
Participants:

 Description   

SERVER-32755 identified several microbenchmarks that will be helpful in monitoring (new) unique index performance.

The new unique index can have old styled entries after an upgrade and any new entries will be made in the new style. Hence with an upgrade mixed entries will be a common scenario. It will be worthwhile adding microbenchmarks for this mode after the new index switches from gating variable to FCV.



 Comments   
Comment by Sulabh Mahajan [ 01/Aug/18 ]

I discussed this task with the team and we have decided not to add this test to microbenchmarks. After going over the implementation of the new unique index, it appears that the logic that takes into account whether or not we have mixed entries is the same as for the pure new index. With that understanding we do not see any benefit of adding tests for the mixed entries.

I am closing this ticket.

Comment by David Daly [ 07/May/18 ]

If you could describe it in javascript, we should be able to do it in microbenchmarks, although we would want to be careful about impact on subsequent tests. There are pre and post hooks available in mongo-perf. For example, see here

Comment by Alexander Gorrod [ 07/May/18 ]

max.hirschhorn or david.daly Could you let us know if there is existing infrastructure that would allow us to run the microbenchmarks on an index that has content created with both old and new format entries? It would probably require an FCV upgrade during the populate or a multi-phase populate using different versions.

Generated at Thu Feb 08 04:33:43 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.