Spot instances

Spot instances are spare capacity that AWS sells at a discount (up to 90%). The caveat is that spot instances may not always be available, and can be recalled by AWS at anytime. Cortex allows you to use spot instances in your cluster to take advantage of the discount while ensuring uptime and reliability of APIs. You can configure your cluster to use spot instances using the configuration below:

# cluster.yaml
# whether to use spot instances in the cluster; spot instances are not guaranteed to be available so please take that into account for production clusters (default: false)
spot: false
# additional instances with identical or better specs than the primary instance type (defaults to only the primary instance)
instance_distribution: # [similar_instance_type_1, similar_instance_type_2]
# minimum number of on demand instances (default: 0)
on_demand_base_capacity: 0
# percentage of on demand instances to use after the on demand base capacity has been met [0, 100] (default: 50)
# note: setting this to 0 may hinder cluster scale up when spot instances are not available
on_demand_percentage_above_base_capacity: 0
# max price for spot instances (default: the on-demand price of the primary instance type)
max_price: # <float>
# number of spot instance pools across which to allocate spot instances [1, 20] (default: number of instances in instance distribution)
instance_pools: 3
# fallback to on-demand instances if spot instances were unable to be allocated (default: true)
on_demand_backup: true

Spot instances are not guaranteed to be available. The chances of getting spot instances can be improved by providing instance_distribution, a list of alternative instance types to the primary instance_type you specified. If left blank, Cortex will only include the primary instance type in the instance_distribution. Cortex defaults the max_price to the on-demand price of the primary instance.

Spot instances can be mixed with on-demand instances by configuring on_demand_base_capacity and on_demand_percentage_above_base_capacity. on_demand_base_capacity enforces the minimum number of nodes that will be fulfilled by on-demand instances as your cluster is scaling up. on_demand_percentage_above_base_capacity defines the percentage of instances that will be on-demand after the base capacity has been fulfilled (the rest being spot instances). instance_pools is the number of pools per availability zone to allocate your instances from. See here for more details.

Even if multiple instances are specified in your instance_distribution on-demand instances are mixed, there is still a possibility of running into scale up issues when attempting to spin up spot instances. Spot instance requests may not be fulfilled for several reasons. Spot instance pricing fluctuates, therefore the max_price may be lower than the current spot pricing rate. Another possibility could be that the availability zones of the cluster ran out of spot instances. on_demand_backup can be used mitigate the impact of unfulfilled spot requests by enabling the cluster to spin up on-demand instances if spot instance requests are not fulfilled within 5 minutes.

There is a spot instance limit associated with your AWS account for each region. You can check your current limit here (set the region in the upper right corner to your desired region, and search for "spot"). Note that the listed spot instance limit may misrepresent the actual number of spot instances you can allocate. Your actual spot instance limit depends on the instance type you have requested. In general, you can run a higher number of smaller instance types, or fewer large instance types. For example, even if the limit shows 20, if you are requesting large instances like p2.xlarge, the actual limit may be lower due to the way AWS calculates this limit. If you are not getting the number of spot instances that you are expecting for your instance type, you can request a limit increase here.

Example spot configuration

Only spot instances with backup

spot: true
on_demand_base_capacity: 0
on_demand_percentage_above_base_capacity: 0
on_demand_backup: true # recommended for production clusters

3 on-demand base capacity with 0% on-demand above base capacity

min_instances: 0
max_instances: 5
spot: true
on_demand_base_capacity: 3
on_demand_percentage_above_base_capacity: 0
# instance 1-3: on-demand
# instance 4-5: spot

0 on-demand base capacity with 50% on-demand above base capacity

min_instances: 0
max_instances: 4
spot: true
on_demand_base_capacity: 0
on_demand_percentage_above_base_capacity: 50
# instance 1: on-demand
# instance 2: spot
# instance 3: on-demand
# instance 4: spot