The use of multiple workers allows for much higher throughput, but it also allows for race conditions during processing. When messages for the same user are sent to different workers, handling order can no longer be guaranteed. Therefore we have failed to fulfil our business rule. If we were able to ensure every message from the same user were sent to the same worker, the worker could ensure those messages are handled in order while the system as a whole would still benefit from the degree of parallelism. But how do we make this happen?
Uptime and SSL certificate monitoring doesn't have to be boring, confusing or expensive. Together with my buddy Mattias Geniar I'm currently building Oh Dear: a webapp to keep an eye on the health of your websites. Register here to try it out. (sponsored link)