[Nomad Autoscaler] 1 time scaling action

Nomad
1

Hi nomad teams…

we have condition like this :
imagine that we have multiple nomad clients provider across global,
lets say we have AWS and Serverscom for now,

as Serverscom didn’t provide elasticity of instance autoscaling

we need to have one time autoscaler policy that scale the AWS Provider from 0 to 1 if serverscom resource out of capacity…

and another autoscaler policy (for aws client autoscaler) if above policy triggered, if not do nothing,

then another policy again to scale aws to 0 if serverscom have many unallocated resource based on threshold,

can I accomplish above with current nomad autoscaler?
maybe using fixed-value strategy with conditional query? any advise or idea how accomplish above problems?

2

Just a little side note:

We created scalr, a generic autoscaler for public/private clouds not offering built-in autoscaling. In case “serverscom” provides an restful API, this wouldn’t be too hard to implement.

3

they didn’t provide it for Dedicated Server… afaik :cry:

4

Solved by using advanced promql query and pass-through strategy,

example query :

(((sum(nomad_client_unallocated_memory{provider=~"serverscom|nitrado", node_status="ready", node_scheduling_eligibility="eligible"}) < bool 2000) == (absent(sum(nomad_client_unallocated_memory{provider="aws", region="us-east-1", node_class="us-east-1-aws-001", node_status="ready", node_scheduling_eligibility="eligible"})) or clamp_max(absent(not_exists),0))) or clamp_max(absent(not_exists),0))

and then we combine with nomad client aws asg config