Currently you collect stats about CPU and RAM usage with a frequency that is too low to see spikes of traffic.
My web service handles bursts of requests that have a duration that might be less than 1 minute.
In that case, even if my servers are under heavy load, and maybe droplet resources are not enough, your graphs show that everything is ok and resource usage is low.
Please collect stats with an higher frequency (e.g. every 10, 20 or at least 30 seconds... 1 minute is too much).
We are currently exploring a paid, advanced monitoring option with configurable sampling rates. We are not planning to increase the metrics sampling rate on default monitoring.
Attachments Open full size