ECS CPU metric inconsistency
gautam-nutalapati opened this issue · comments
ecs.task.cpu.utilized and ecs.task.cpu.reserved metrics units are confusing
We upgraded to aws-otel-collector:v0.35.0
. Our graphs became inconsistent after we upgraded.
Simplifying my previous question:
Can we add units to ecs_task_cpu_utilized_None
and ecs_task_cpu_reserved_None
metrics? Similar to how new version on library adds units to other metrics generic_counter_short_total
?
My observation is below based on numbers I have been seeing in our graphs:
If service is on EC2 backed ECS, ecs_task_cpu_reserved_None is in vCPU.
If service is on EC2 backed ECS, ecs_task_cpu_utilized_None is in vCPU units.
If service is on Fargate backed ECS, ecs_task_cpu_reserved_None is in vCPU units.
If service is on Fargate backed ECS, ecs_task_cpu_utilized_None is in vCPU units..
Can we confirm if these units are correct meanwhile?
Old post:
Before upgrade,ecs_task_cpu_utilized_None
reported metric in percentage.
After using this version, this metric is now reporting CPU utilization in CPU units. (1vCPU=1000 units)
Documentation still mentionsecs.task.cpu.utilized
is in percentage.Update:
If service is on EC2 backed ECS,ecs.task.cpu.utilized
is in vCPU units.
If service is on Fargate backed ECS,ecs.task.cpu.utilized
is in percentage.
Steps to reproduce if your question is related to an action
Deploy ECS task using aws-otel-collector:v0.35.0
version of aws-otel-collector.
What did you expect to see?
ecs_task_cpu_utilized_None
and ecs_task_cpu_reserved_None
metrics should have units in metric names.
I expect ecs_task_cpu_utilized_None
reports metric in percentage but its reporting in vCPU units.
Environment
N/A
Additional context
N/A
This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 30 days.
This issue was closed because it has been marked as stale for 30 days with no activity.