brenmcnamara / infindi-backend Goto Github PK

Does flow-typed have firebase types?

If not, should create some automatic way of referencing firebase types for project

Adding logging module for web and worker services, should use this during initialization

• Need a way to quickly know that job requests are being listened to by worker

• Should be able to send a ping from the command line that gets a response from a worker if it's listening

• Ping from command line

• Worker detects the ping and just marks the job as complete without doing anything else

• Detect the change in the job we sent and if it gets mark as complete, then we are golden!!

• Currently, when creating a new request, we check if the device id exists for an open request

• Should change this check for device id and bundle identifier combination so that we can have multiple apps with sessions and we can have a debug and production build with open sessions

• If there are more than X number of open requests with the same device id and bundle identifier, don't let user create a new one

  • Need to create some buffer for abandoned sessions to get picked up by the session cleanup job. This will probably start-off running about once an hour

When creating a new session, we need to check for non-responsive sessions with the same device id and do something about them.

Reset net worth when deleting accounts

replace needs regex that is globally applied

• Need to know precise payload that is getting passed to the webhook every time

• Need to know the IP Address it's coming from

• Need to see the headers

Express responses return with a header "X-Powered-By": "express". We don't want people knowing what we are using in the backend, since it may give them some clues on how to exploit the system. This header should not be returned.

• It's a pain in the ass to run react native when I want a new item id
• There should be a web client to quickly do this

• Each service should have structured and detailed logs
• Logs should include severity (info, warning, error), topics, routes, etc...

After deploying our services, we need a way to make sure that download requests are being worked on.

One solution is to write a quick script that detects changes to download requests and reports status updates:

(<id 1>): (null) -> NOT_INITIALIZED
(<id 2>): (null) -> NOT_INITIALIZED
(<id 1>): NOT_INITIALIZED -> IN_PROGRESS
...

Cleanup and handle abandoned jobs

Calling update all has a couple of caveats:

1. Should NEVER be run alongside anything that is writing to user data, there can be race conditions. Need a good solution for this.
2. Due to Firebase batch limit of 500 updates, this can result in an error if trying to update too much data. (Can fix this by running a set of batches instead of a single batch)

Download requests should be cancel-able

• Should send the job to update the metrics when accounts are added via plaid.

• Should have some load indicator to show that metrics are in the middle of an update. In the data model, should just have a flag to indicate if a job is in the middle of updating the metrics.

Middleware utils is being duplicated in worker and web, should migrate this to common-backend

We are currently copying that repo between mobile and backend. Should make it its own repo and import it via package manager

Some runtime errors that happen when a job is running result in a crash without getting caught and handled by the error-handling code. Need to figure out why this happens and fix it.

When attempting to login with a user, we replace the current yodlee provider account with the next one. The problem is that if the login is a failure, then we fail to login. I also suspect we will lose transaction data, even on a successful login, need to think through the user flow for this.