Spot fleet request over-provisioned

Question

Hello. I'm having a spot fleet request that is using launch template to deploy DEV/QA environment into Rancher. Lately it has been over-provisioned, meaning that total capacity is set to 6 for DEV and 5 for QA, however im having up to 10 instances at certain moments in DEV and 8 instances in QA. What i have noticed is that instance weight has been set to 1 for every instance type. MY instance fleet is r5.xlarge weight=1, r5d.xlarge weight=1, r4.xlarge weight=1, m5.2xlarge weight=1 and i don't see in the UI where you can set nor modify instance weight upon creation of the spot request nor afterwards. Also when you create spot request without template, instance weight is not applied. So since we don't need 10 instances running my question is how can i limit that to 6 as stated in the target capacity and not having resources over provisioned? Thanks in advance

Accepted Answer

Hello, thank you for your post. There are scenarios when the number of running instances will be higher than the targeted capacity. For example, if you are using Spot Fleet Capacity Rebalancing[1] and the instance replacement strategy is "Launch only" then the old instances will remain running until you terminate them manually. In order to better understand your existing configuration and determine if you are observing intended behavior, or an unintended problem, I highly recommend you open a support case so that our engineers can take a closer look.

References:
[1] https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/spot-fleet-capacity-rebalance.html

References:  
$$1$$ https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/spot-fleet-capacity-rebalance.html

Answer

Much appreciated for taking the time to answer my question. Yes so originally launch only was set so that might be the reason for this behaviour.

Spot fleet request over-provisioned

Contenido relevante