Controls how a template is picked when several AMI configurations in this cloud match the same label.

Unticked (the default), the templates are tried in the order they are configured, which is the historical behaviour. If the first one cannot provision, the next one is tried for that request, but the next request starts from the first template again.

Ticked, all templates matching a label are treated as interchangeable hardware — the admin gave them the same label, so any of them can run that work — and Jenkins rotates between them. The rotation honours each template's Hot spare weight, so a spot template with weight 3 is attempted three times as often as an equivalent on-demand template with weight 1, while still falling back to on-demand immediately when spot capacity is unavailable.

With rotation enabled, a template that reports insufficient capacity is also demoted for a few minutes so subsequent requests for that label go straight to a template that can supply the instance type. This demotion never applies to a label served by a single template, which keeps retrying across its own availability zones.

Weights are shares of the requests, so every weight keeps receiving some of the work. If they should instead rank the templates, so that a lower weight is only used once the higher one is exhausted, see Saturate the highest weight before using a lower one.