[SERVER-47363] Define correct use of MemberIds in Atlas workflow tests Created: 06/Apr/20  Updated: 29/Oct/23  Resolved: 03/Jul/20

Status: Closed
Project: Core Server
Component/s: Replication
Affects Version/s: None
Fix Version/s: 4.7.0

Type: Task Priority: Major - P3
Reporter: Judah Schvimer Assignee: Suganthi Mani
Resolution: Fixed Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Depends
Documented
is documented by DOCS-13745 Investigate changes in SERVER-47363: ... Closed
Related
related to SERVER-54746 Two primaries in a replica set can sa... Closed
related to SERVER-48345 Increase maximum memberId in a replic... Closed
related to SERVER-49938 Create Atlas workflow tests of what A... Closed
is related to SERVER-47852 Two primaries can satisfy write conce... Closed
is related to SERVER-54746 Two primaries in a replica set can sa... Closed
Backwards Compatibility: Fully Compatible
Sprint: Repl 2020-06-29, Repl 2020-07-13
Participants:

 Description   

MemberIds should uniquely identify nodes, even across config versions. Therefore reusing MemberIds can be dangerous. The current Atlas workflow tests will do whatever Atlas does today, but we should decide what the correct workflow is and change the tests to match that so Atlas can implement the correct workflow.



 Comments   
Comment by Githook User [ 03/Jul/20 ]

Author:

{'name': 'Suganthi Mani', 'email': 'suganthi.mani@mongodb.com', 'username': 'smani87'}

Message: SERVER-47363 Defines correct use of MemberIds in Atlas workflow tests.
Branch: master
https://github.com/mongodb/mongo/commit/eeda016f9579c9e880b7bc2a8d0d7e76501c88f2

Comment by Judah Schvimer [ 27/May/20 ]

I think that is the safest recommendation.

Comment by William Schultz (Inactive) [ 27/May/20 ]

To clarify, going forward will we be recommending that MemberId be effectively treated as a globally (in time and space) unique id (e.g. a UUID)?

Comment by Judah Schvimer [ 20/May/20 ]

Yes, I agree with the above, and think it's a good model for thinking about nodes and member ids.

Comment by William Schultz (Inactive) [ 20/May/20 ]

judah.schvimer Do you agree that a "node" as used in the ticket description, should, for the most part, correspond to a particular set of data files? For example, as you mentioned elsewhere, if you want to resync a node, you could remove the node first, and then add back a new node with a new, unused member id. This would uphold the assumption that each member id maps to a unique set of data files.

Comment by Judah Schvimer [ 01/May/20 ]

We should also document this more widely for all users in our documentation.

Generated at Thu Feb 08 05:13:57 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.