Distributed counters for a programming competition
requires Python3.3+ and MongoDB
For automatic failover to work at least 3 MongoDB instances should run together in a replica set.
See startmongo.sh script for an example how to start 3 MongoDBs.
You start application server with python zemcounters/server.py. This will start a process for each core your computer has.
If you want more than one application server you need to set up a load balancer (preferably nginx) infront.
API:
/cnt/<collection>/
regex for collection: ([\w]{1,128})
POST: Creates a counter in a collection and returns the location in Location header with status 201.
MongoDB's namespace limitations apply here, so you are limited in amount of collections you can have.
/cnt/<collection>/<counter_id>
regex for counter_id: ([a-zA-Z0-9]{24})
GET: Returns the value of the counter via JSON.
POST: Increments the counter by 1.
DELETE: Deletes the counter. Returns 1 via JSON if counter was deleted and 0 if there was no counter with this ID to be deleted. This doesn't mean an error because if network error occurs counter can be deleted and then on retry 0 is returned.
/cnt/<collection>/<counter_id>/<n>
POST: Increments the counter by n.
/cnt/<collection>/<counter_id>/reset
POST: Resets the counter to 0.
TAILER:
If you need to poll counters constantly please use tailer because it scales so much better.
You can access tailer via ws://hostname/tail/ you can optionally include counter id /tail/<counter_id>
Once connected you can subscribe and un subscribe by sending json via websocket. {'<counter_id>': 's'} subscribes you to the counter and {'<counter_id>': 'u'} un subscribes.
When counters are updated you get messages in format: {'id': '<counter_id>', 'd': 1} if the counter was deleted and {'id': '<counter_id>', 'n': n} with n being the new value of the counter.