trivago / Scalad

Env variables:

• PORT = "80": Port on which the application is going to be listening.
• NOMAD_HOST = "http://nomad.service.consul:4646": Nomad host endpoint.
• NOMAD_REGION = "global": Nomad Region.
• VAULT_TOKEN = "ljkasdflkjafd": Token for vault with permission to access every secret needed for all the scale jobs.
• USE_SLACK = "true": Flag to enable the use of slack as a system message application.
• HTTP_USER = "user": User needed for basic auth for endpoints to scale up or down manually.
• HTTP_PASS = "password": Password needed for basic auth for endpoints to scale up or down manually
• METRICS_ENDPOINT = "http://prometheus.yourorg.com/api/v1/query?query=": Endpoint from where to get the metrics which are going to be used to triger the scale events.

Inside Job file (At taskGroup lvl):

meta {
    scaler = "true" Activate the scaler
    min_query = "sum(rate(nomad_client_allocs_cpu_total_ticks{exported_job='scaler-test'}[1m]))by(exported_job) < bool 1" Query that gives the Min threshold for scaling down
    max_query = "sum(rate(nomad_client_allocs_cpu_total_ticks{exported_job='scaler-test'}[1m]))by(exported_job) > bool 2" Query that gives the Max threshold for scaling up
    query_fire_time = "2m" Time the query need to be true before triggering the saling event
    scale_cooldown_down = "20s" Time in cooldown for a scale dow event
    scale_cooldown_up = "25s" Time in cooldown for a scale up event
    scale_count_down = "1" Amount of containers that are going to be removed on a scale down event
    scale_count_up = "2" Amount of containers that are going to be added on a scale up event
    scale_max = "16" Maximun amount of containers
    scale_min = "1" Minimun amount amount of containers
}