918625 - Implement a distributed mutex for bug modifications

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Gaia::GithubBot, defect)

Description

[John Ford [:jhford] CET/CEST Berlin Time]

Updating bugs (i.e. more than just comments) requires two non-atomic API calls.  We need to fetch the bug to get the update token, then we post the new bug information.

We should have a shared mutex across the entire bot to ensure that no single bug is attempted to be worked on by two different api consumers.

Comment 1

[John Ford [:jhford] CET/CEST Berlin Time]

Attachment: pull request

This is a distributed lock.  It's safe to double-unlock something, but double locking will fail (obviously).

I'm not sure if I like how the redisOpts stuff works, also I think my unit tests would be better if I stubbed out redis instead of using a real redis.  I didn't do that because I couldn't figure out how to stub api methods without recreating a whole api object.

Comment 2

[James Lal [:lightsofapollo]]

Comment on attachment 811415 [details] [review]
pull request

with nits see github

Comment 3

[James Lal [:lightsofapollo]]

https://github.com/mozilla-b2g-bot/shared-mutex/commit/e62dcf62b0b4d84b2c1f92e2726a519de6203b3c