The autoscaling features are controlled via the VIRTUAL_ASG_MAX_INCREASE (default = 10 instances) and VIRTUAL_SCALING_INTERVAL (default = 30s) config options, which control the amount (number of instances) and interval (in seconds) of allowed increases. This puts a limit on the rate at which the ASG's desired_size can be increased.
Also added metrics to the scheduler, to display these quantities.
Also added request timing metrics to the scheduler. These are showcased in the "Serratus Scheduler" Grafana dashboard. With these, we should be able to get a better idea of if / where our scheduler is hanging up, by how much, and where to focus improvements.
The autoscaling features are controlled via the
VIRTUAL_ASG_MAX_INCREASE(default = 10 instances) andVIRTUAL_SCALING_INTERVAL(default = 30s) config options, which control the amount (number of instances) and interval (in seconds) of allowed increases. This puts a limit on the rate at which the ASG's desired_size can be increased.Also added metrics to the scheduler, to display these quantities.
Also added request timing metrics to the scheduler. These are showcased in the "Serratus Scheduler" Grafana dashboard. With these, we should be able to get a better idea of if / where our scheduler is hanging up, by how much, and where to focus improvements.