Wave deployment with parallel actions

The parallel option -p is available for all commands (e.g. start/stop/update/lock/...) acting on multiple components:

• -p <p-spec> parallel execution according to p-spec

Formats of the p-spec:

• -p <number> use number parallel workers for each plan (even nested ones)

  example: start all services with 10 parallel workers

  start service://* -p 10

• -p "<plan>=<items>_<workers>_<errors>"

  • plan: name of the subplan (use -n to see the actionplan)
  • items: number of items in the given plan to apply the following parameters
  • workers: number of parallel executions on given items
  • errors: tolerated errors during execution
  • an asterisk * translates to up to all remaining items

  example: during an update, start all unaffected services as fast as possible (or: in parallel)

  update -n -p "update/prestart=*_*_0"

  Note: only the prestart plan gets executed in parallel, all other plans (notably the /update/stopupdate/start plan) gets executed sequentially.

• -p "<plan>=<items>_<workers>_<errors>[:<items>_<workers>_<errors>[...]] [<plan>=...]"

  • more complex specification
  • splits a plan in severall parallel plans, configures multiple plans

• Complex Example: Pilot Phase Updates

  • try an update on the first chunk, fail on any error
  • try an update on the next 5 chunks with 2 workers, fail on 2nd error
  • update all remaining chunks with 10 workers, fail on an error rate >10%

  update/stopupdatestart=1_1_0:5_2_1:*_10_10%"```