clustered-cron-example is a proof of concept on how to execute a cron job in a multihost environment without sacrifying availability or firing more than one job at a scheduled time.
It is based on the amazing real-time web framework Mojolicious, and the resilient etcd distributed database system.
- Docker (v19+), Docker compose (1.27+)
- Haven't checked, but as the project runs 3 containers at the same time, you should have a generous amount of RAM available (>= 8GB I guess)
- git clone the project and cd to it, then run this command:
docker-compose up- after downloading, building and starting, you should see that etcd initializes ok with 3 nodes.
- then you should receive 1 warn every 10 seconds, with the simulated Cron job results
- the crontab is fired every 10 seconds, but this is only to show quick results. In practice, when a node goes down, it is important that the rest of the etcd nodes stabilize as soon as possible. One scheduled time while it is stabilizing should be fine, but I don't know if you lower this interval too much.
- to make any changes and restart, you need to stop the containers (^C), then run
docker-compose down
docker-compose up --build- You should have 3 nodes running. The corresponding web services would be at localhost ports 9010, 9011 and 9012. They just show a "Hello World" message when you point a browser to them, just to show they are alive. (The actual interesting stuff is what you get into the logs).
- The nodes are: web_0, web_1, and web_2
- from other terminal, you can stop one of the modules (for instance web_2):
docker stop clustered-cron-example_web_2_1- then you can start it again:
docker start clustered-cron-example_web_2_1- The first host that gets the exclusive lock of the next scheduled time runs the task.
- That doesn't change very often, but you will see changes after several minutes.
- You should be able to enter to any of the nodes (for instance to web_0 node), from a different terminal:
docker exec -it clustered-cron-example_web_0_1 bash