The converge function can be used to adjust node counts and to apply phases. The :bootstrap phase is the first phase applied to any new node that is started, and :configure is always applied. Additional phases may also be specified. If the :configure phase is not explicitly passed, then it will always be applied as the first phase (or second, after :bootstrap on new nodes).

In this example we define a function that changes the number of running nodes for the "mygroup" group.

  (require 'pallet.core)

  (defn scale-cluster [n]
      (pallet.core/group-spec "mygroup" :count n)
      :compute (pallet.compute/service "aws")))

converge also accepts a :prefix keyword argument, which is applied to the group names in the call. This can be used to build job specific clusters. In this example we scale load-balancer, web app and database nodes using a single load balancer and twice as many web app frontends as database backends.

  (require 'pallet.core)
  (def load-balancer (pallet.core/group-spec "lb"))
  (def web-app (pallet.core/group-spec "webapp"))
  (def database (pallet.core/group-spec "db"))

  (defn scale-cluster [prefix n]
       {load-balancer 1
        web-app n
        database (inc (/ n 2))}
       :prefix prefix
       :compute (pallet.compute/service "aws")))


The lift function is used to apply phases but does not change node counts. The :configure phase is run by default only if no phases are explicitly specified.