[SERVER-37672] Re-enable agg fuzzer on debug builds Created: 19/Oct/18  Updated: 06/Dec/22  Resolved: 14/Aug/19

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

Type: Task Priority: Major - P3
Reporter: Robert Guo (Inactive) Assignee: Backlog - Server Tooling and Methods (STM) (Inactive)
Resolution: Won't Fix Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Related
related to SERVER-42616 $stdDev accumulator can return imprec... Backlog
related to SERVER-37668 Disable the aggregation fuzzer on Win... Closed
is related to SERVER-37429 $avg of Infinity and null is NaN on d... Backlog
Assigned Teams:
Server Tooling & Methods
Participants:
Linked BF Score: 56

 Description   

Re-enable the fuzzer on Windows DEBUG and Coverage builds.



 Comments   
Comment by Max Hirschhorn [ 14/Aug/19 ]

The $group stage spills all documents to disk when running on a --dbg=on build. This is problematic for the aggregation fuzzer because it can lead to documents being processed in a different order than the original document source. Due to commutativity issues in $group's accumulators, the aggregation fuzzer may therefore see different results based on whether it spilled to disk or not.

Addressing these kinds of issues (e.g. SERVER-37429 and SERVER-42616) isn't a priority for the Query team right now. In the future, we could consider adding a mode for the aggregation fuzzer where it compares the behavior between when spilling to disk and when not spilling to disk. Based on conversations with charlie.swanson, we would want to do this by introducing a failpoint to enable always spilling to disk so we aren't implicitly testing this behavior by downloading a --dbg=off last-stable binary when running multiversion tests on a --dbg=on builder.

Generated at Thu Feb 08 04:46:40 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.