Project 7 Changes and Updates

New features since release 0.0.7

The following features are made available in project 7:

1. Added spark-based pet recommendation system
2. Added an autocomplete search bar with search history metrics
3. Added an advanced fixture to automate SQL population / Elasticsearch indexing of placeholder pets / views

The following features are deprecated in project 7:

1. Removed jmeter performance testing to avoid 503 error on low-grade hardware
2. Since performance testing is removed, load balancing module based on haproxy was also removed from docker-compose to avoid 503 error on low-grade hardware

Portia: user stories (continuously updated)

1. As the seller, I want to inform the customer what type of animal does the pet belongs to
2. As the seller, I want to update the information of the pet to let the customer know the most up-to-date condition of the pet
3. As the seller, I want to request to cancel sales when the pet is no longer available
4. As the customer, I want to see all the pets listed by all the sellers
5. As the customer, I want to type and search about the pet I want
6. As the customer, I want to change and update my profile to give my most up-to-date information to the seller
7. As the seller, I want to be able to log in with my account, so I can ensure nobody else else can edit information of my pets without my permission
8. As the seller, I want to be able to log out with my account. If I log in to the website in a computer in public, I want to ensure nobody else can use my account after I leave and am no longer using that computer
9. As the customer, I want to see all pets specified by a specific seller because I really all the pets in his / her inventory and want to follow that seller
10. As the seller, I want to be able to reset my account password so I can log in and retrieve my inventory information if I happened to forget the account password

Deployment & Testing

Suggested testing workflow

1. Ensure the init script for each container is executable by running the following command:

   $ sudo chmod -R 777 internet-scale-app_00X
   

2. Ensure that a mysql container with a database named cs4260 and a user 'www'@'%' who is granted all privileges to cs4260 and test_cs4260 (the test database Django test Client will create later). Otherwise, docker-compose up will not bring up any container

3. Add mysql container to docker networks backend by running the following command:

   $ sudo docker network connect internet-scale-app_backend mysql
   

   Depending on your docker version, network may be named to internetscaleapp_backend

4. Run sudo docker-compose up in project root folder to bring up docker containers

   Note: it is possible that spark-worker will try to connect to spark-master multiple times before it forms a cluster with it. This is OK since spark (and most other distributed computing frameworks) will always attempt connecting until it forms a cluster.

5. Head to localhost:8006/homepage to access the project:

   • After docker-compose is fully started, run the advanced fixture with the following command:

   $ sudo docker exec -it presentation /bin/bash
   $ python fixture.py
   

   If successful, running the fixture will produce the following output:

   

   The fixture will register a user with username test_user and password ABC123456789 and will then create 3 pets. After that, it will view each pet 5 times (cache-free)

   • Run spark script to start recommendation service by running the following command (in a separate terminal):

   $ sudo chmod 777 data/update_recommendation.sh
   $ ./data/update_recommendation.sh
   

   Recommendations will be updated every 60 seconds

   • To create a new pet, click [Register] to register as a new user first
   • After registration, user will be redirected to login page. Click [Log in] after filling in user credentials. A user who is already logged in will be redirected to homepage
   • After logging in, click [Create a new pet!] to create a new pet
   • After a new pet is created, user will be redirected to homepage
   • Click [Check it out!] on each pet created to view its detailed information
   • Since pet details are cached in redis container (cache will be invalidated every 20 minutes / when a user logs out). When a pet is cached, accessing it detail page will not be logged (as a result, its recommendation feed will not be updated either). Therefore, to test out the recommendation service, it is recommended to create multiple users / log out and re-log in again between viewing each pet. Only co-viewed pet with 3 or more views will be considered as a recommendation. Recommendations are updated every 60 seconds by running update_recommendation.sh. An example of recommendation feed produced from pre-loaded fixture data looks like as follows:

   

   • To test out the updated search bar, simply type in it: suggestions from search history will appear (if the user has one) as follows:

   Note: the suggestion panel will not show if input does not match any item from search history or when it is the first time a user searches (i.e. when there is no search history)

   • If the suggestion panel is blocking any content on the page (for example: the login and register page), simply click outside of the suggestion panel to clear it