[SERVER-32558] Consider removing the sharding continuous stepdown suite Created: 05/Jan/18 Updated: 23/Sep/18 Resolved: 27/Aug/18 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | Sharding, Testing Infrastructure |
| Affects Version/s: | None |
| Fix Version/s: | None |
| Type: | Task | Priority: | Major - P3 |
| Reporter: | Kaloian Manassiev | Assignee: | Max Hirschhorn |
| Resolution: | Won't Fix | Votes: | 0 |
| Labels: | None | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Participants: |
| Description |
|
The sharding continuous stepdown suite was introduced in MongoDB 3.2 as a way to test CSRS and over time has surfaced numerous sharding and replication bugs. At that time, the sharding metadata commands were all implemented on mongos and were doing reads and writes against the config server. As of MongoDB 3.6 however, most of the metadata commands have been moved to run on the config server primary. As a result of this, the continuous stepdown suite has now become a test of the repeatability of these operations in the context of running sharding tests. However, due to the different command return values which can happen after an operation is repeated, this suite is now generating build failures, which are not necessarily bugs, but behaviour, which is not expected by tests and making tests account for this behaviour is not a trivial task, nor it is something which has a great customer value. Because of this I would like to propose that we get rid of the sharding continuous stepdown suite in lieu of the already running concurrency stepdown suite. |
| Comments |
| Comment by Kaloian Manassiev [ 27/Aug/18 ] |
|
Like Esha mentions, with the recent changes to push some of the ManualInterventionRequired operations further out to the shards, the suite no longer generates nearly as many known BFs as it did right after we pushed logic from mongos to the config server. Because of this I am not that eager to get rid of it yet, since there is some value in testing the config server stepdowns. Because of this I am fine with closing this ticket for now. |
| Comment by Esha Maharishi (Inactive) [ 06/Aug/18 ] |
|
max.hirschhorn, I agree the config stepdown suite was greener before the "move metadata commands" project. I think this is not because these issues didn't exist for a user, but because the issues were not revealed by our test infrastructure, since we have never had a "mongos restart suite." I personally think we should keep the config stepdown suite, given the sharding team's recent changes to initial split and planned work on the sharding catalog. |
| Comment by Gregory McKeon (Inactive) [ 26/Jul/18 ] |
|
max.hirschhorn kaloian.manassiev can we revisit this, particularly in light of the bf's linked to |