We use Datadog for all our alerts. I’m hoping to recreate an alert for each of our apps off of a chart that is pretty much the equivalent of the CPU usage chart I see in the Aptible dashboard. However, the only metric that is readily available to me (as far as I can tell) is the Enclave’s milli CPU usage. As I understand it, that tells me how many thousandths of a core of the host’s CPU I’m using. However, I haven’t been able to see anywhere how many cores our hosts use. Is there a reliable way I can work that out? Failing that, is there another approach you think I should follow to monitor CPU usage?
It’s currently on our team’s radar to improve the CPU usage metrics that we emit but we don’t have a timeline for that update just yet. In the meantime, you can use the following formulas to determine the allocated CPU for each container profile based on the enclave.memory_limit_mb :
General (¼ CPU: 1GB RAM):
milli_cpu_allocated = enclave.memory_limit_mb / 1024 / 4.0 * 1000
RAM-Optimized (⅛ CPU: 1GB RAM):
milli_cpu_allocated = enclave.memory_limit_mb / 1024 / 8.0 * 1000
CPU-Optimized (½ CPU: 1GB RAM):
milli_cpu_allocated = enclave.memory_limit_mb / 1024 / 2.0 * 1000