I'm looking for a good way to monitor concurrent users in my server. The server's got a websocket connection, so it's easy to determine whether a user is connected or not. After looking into CloudWatch for a bit, I have a plan:
Every minute, each server in the server cluster will send the current amount of users connected to the same CloudWatch metric and namespace
In the end, I can use the CloudWatch Sum view to see the concurrent users in that period - provided that I use the 1 minute period. I don't have a lot of experience with CloudWatch, and that makes me wonder if that's a good approach. I know even for graph visualizations, there are limits on the amount of information that CloudWatch can display at a time. I know too that I could make a CW graph that could e.g. use the sum for a 15-minute period, and divide that by 15, if I wanted to see higher time frames. But that makes me wonder if I'm misusing CloudWatch, if that plan makes sense and what kinds of issues I can run into that I haven't foreseen.
For example, will I be able to add an anomaly alert, if the data is structured in this manner? Is there a better way to achieve that? Does this plan make sense?
Thank you!
