Auction system which will allow users to concurrently bid on items for sale.

PS.: This auction allows the user to bid with a somaller value than the last one. =D

Each of the container above has a propose.

• Locust: Simulate user's bids and execute a load testing to measure performance of Bid API
• Auction: Responsible for auction's endpoints such as
• Bid: RestAPI to interacte (create, update, delete, get, list) with bids
• APM elastic/Elasticsearch/Kibana: Implement APM to measure performance on aplications

Items quantity: Set the quantity in docker-compose file, line 26

$ git clone https://github.com/tuliocpbs/auction-bid-tracker.git

$ cd auction-bid-tracker

$ docker-compose up -d

Wait for all dependencies be up

Dependes on your docker settings, you will need to set more memory so the elasticsearch stay up. For that you can use:

• For linux SO: $ sudo sysctl -w vm.max_map_count=262144

• For mac OS: Open Docker Desktop -> Select Resources -> Set more memory

The datas structures used are:

API responsible for provide information about the auction, such as

Access Auction's swagger doc : http://localhost:5000/docs

To user any requests insert the Api-Key value on Authorize field

Endpoints available:

• /api/v1/bidsofanitem (GET)
• /api/v1/currentlywinnig (GET)
• /api/v1/itemswithuserbids (GET)

API responsible for create, delete and provide information about bids.

Access Bid's swagger doc : http://localhost:5001/docs

To user any requests insert the Api-Key value on Authorize field

Endpoints available:

• /v1/bid (GET/DELETE/PUT)
• /v1/bids (GET)

Wait for Kibana be up

Test the performance of Bid RestAPI with Locust

In this project to deal with the concurrency caused by many users doing bids at the same time, was necessary create a lock on bids variable. I used a solution provide by Python's standart library, threading.

This example explains why this concurrency problem happens:

1. Two users (Ross and Chandler) try to do a bid on the same item (jewel) and in the same time.
2. To different threads get the all bids on this item.
3. The first thread (Ross's bid) update bids variable with Ross's bid.
4. The second thread (Chandler's bid) update bids variable with Chandler's bid, but based on the old data that bids contains, and this data is all bids without Ross's bid. Because of that, the aplication loses Ross's bid.