Blurber

In the Wild

This app can also be found deployed on heroku, at https://dist-blurber.herokuapp.com/

The Front-End server (everything in the main.go hierarchy) is hosted on heroku, and is set up to continously build on the /release branch on our github repo.

The front-end server connects to a backend server hosted by Linode.com (it's cheap).

The three (3!) backend services are hosted on the Linode server, which the heroku app sends gRPC messages to.

Presentation here

Main features

• Login/ register
• Delete account
• Write blurb
• Subscribe to blurbers
• Unsubscribe on feed
• Simple profile page
• Delete personal single blurbs
• OPTIONAL: Feed option to show all blurbs ('Discover' page) vs. just subscription

Roadmap for detaching backend

• Define protobufs for each backend service:

  • BlurbDB
  • UserDB
  • SubscriptionDB

• Split off backends to be behind a GRPC call:

  • BlurbDB
  • UserDB
  • SubscriptionDB

• Ensure that backends are safe to concurrent access from multiple frontends

  • BlurbDB
  • UserDB
  • SubscriptionDB

These will be managed in branches:

1. protobufs
2. split
   • split-blurb
   • split-user
   • split-subscription
3. safe
   • safe-blurb
   • safe-user
   • safe-subscription

Roadmap for replication

1. Repurpose code from Lab 2 to work in this project
   • Replace labrpc code with gRPC
     • protobuf spec
   • Add service/leader discovery to the protocol, instead of static configurations
   • Write unit tests demonstrating functional replication
   • Add a buffered channel for exporting newly committed commands, to be applied elsewhere
     • Integrate this feature with the recovery calls
2. Graft replication daemon onto subscription database, so that it makes use of replication
   • Background thread for processing the buffered commit channel
   • No state changes outside of background thread
   • Unit tests proving it all works
3. Node failure
   • Proving failover security at the daemon level, with unit tests proving efficacy
   • Proving failover security at the backend-tier level, with unit tests proving efficacy
   • Integrating leader failover security into the front-end, maintaining a peers list of backend services, to pivot between on lack of contact

How to run

1. cd your way into ./ and exec ./spinBackend.sh to spin out all the non-frontend servers
2. cd your way into ./ and exec .runLocal.sh, if your FE server is on your lapto. Otherwise, exec ./run.sh if the backend is on the linode server.
3. Point your favorite browser @ http://localhost:4000 if the FE server is running locally, or http://172.104.24.141:4000 if the FE server is running on linode
4. When you're done playing around, exec ./killBackend.sh

Team

• Adam Sanghera (as8513)
• Hieu Do (hsd258)