TRITON-2316: mahi v1 crash on bad role mgmt log entry

Details

Issue Type:Bug
Priority:5 - Low
Status:Open
Created at:2022-07-12T14:58:19.983Z
Updated at:2022-08-24T21:19:03.494Z

People

Created by:Dan McDonald
Reported by:Dan McDonald

Related Links

Description

Reported by Alex Wilson here: [https://github.com/TritonDataCenter/mahi/issues/19|https://github.com/TritonDataCenter/mahi/issues/19|smart-link]
Fixed by Alex Wilson here: [https://github.com/TritonDataCenter/mahi/pull/20|https://github.com/TritonDataCenter/mahi/pull/20|smart-link]

Comments

Comment by Dan McDonald
Created at 2022-07-12T15:01:03.336Z

Description from Alex:
UFDS can write log entries out of order, resulting in a log that can't be replayed by mahi. It's fine for mahi to stop replicating in this situation, but it should not crash.

Currently if you quickly add a user to a role and remove them, and the log entries get re-ordered, then mahi v1 will crash in auth_cache.js around line 738. It should either truck on (because mahi v1 is loosey goosey about that generally) or stop replicating, not crash.


Comment by Dan McDonald
Created at 2022-07-12T15:01:09.372Z

Testing notes from Alex:

testing done:


Comment by Dan McDonald
Created at 2022-08-24T19:14:10.704Z
Updated at 2022-08-24T19:15:29.583Z

https://github.com/TritonDataCenter/mahi/pull/21 Is what we’ll be pulling this in from.


Comment by Jira Bot
Created at 2022-08-24T21:12:33.195Z

mahi commit bcc382b49094144f9fd93f5dff41e57cd0630afb (branch master, by Alex Wilson)

TRITON-2316 mahi v1 crash on bad role mgmt log entry
Reviewed by: Dan McDonald <danmcd@mnx.io>
Reviewed by: Brian Bennett <brian.bennett@mnx.io>