To deal with re-classified URLs, we need a way to queue up actions for various Bouncer instances to take, and a way to update the queue as the actions are resolved or not.
This relates to #96 and #93 and even #92
The flow would be:
BE sends a deferred action to Bouncer
Bouncer attempts the action:
Happy path: The relevant message IDs with the re-classified URL is updated, or
Fail path: The message was deleted, the message was edited to remove the URL, no access, ...?
Bouncer sends the result to BE
BE updates the queue of deferred actions as appropriate
BE records:
The length of time between re-classification and the message ID being updated (happy path), or
BE records failure
Proposed: Time between re-classification and update will be one of our KPIs.
We track mean/median/mode
We also track the bottom 1% performance to see what the worst-hit users and servers experience
If it's bad, we'll have to find a better way to update messages after URLs are re-classified.
To deal with re-classified URLs, we need a way to queue up actions for various Bouncer instances to take, and a way to update the queue as the actions are resolved or not.
The flow would be:
Proposed: Time between re-classification and update will be one of our KPIs.