rafalwilinski / deploy-with-serverless Goto Github PK

Ping me for access to current diagram on draw.io

Add dynamodb-scaling plugin in order to guarantee fast response time when handling more requests

It appears that an attacker could cause your batch job to run arbitrary code in your account by specifying the URL of a github project that they own, or by specifying commands that they want to run in the BEFORE_CMD, PACKAGE_CMD, or AFTER_CMD.

This is a great idea for a service, but it would be safer if you set up a registry of the packages that you want to make available and then allow people to reference packages by ID instead of providing this kind of arbitrary execution power.

Every N image-proxy invocations, there should be version check (build version vs available version). On version mismatch, new build should be triggered (via handler function).

href and image-proxy functions interact on each call with DynamoDB by performing a get request. In order to minimize amount of requests to Dynamo, functions should implement /tmp based cache.

Right now, handler has http event handler. It should be removed and changed to direct invocation.

Before spinning up a machine, validate whether this task was submitted with correct parameters, that includes:

• Correct URL
• GitHub.com as domain
• has package.json
• Serverless framework as dependency

5f9772f#commitcomment-25134485

Just like @alexcasalboni suggested 🙇

Right now, e.g. table names are defined in two separate places: in scripts and in Serverless.yml. it should be defined in one place.

image-proxy function should check build status (whether it's present in dynamodb) and return correct image.

If entry is not found in DB, lambda should return quite different image with some indication that build is in progress and trigger handler function which submits build job.

It should also leverage #2

Status reason: Essential container in task exited
Container message: CannotPullContainerError: API error (404): repository serverless-batch not found: does not exist or no pull access

ef72fa6#commitcomment-25134028

Once again, thanks @alexcasalboni!