Closed jlin21 closed 8 years ago
I hope we're not waiting for each finish their tasks. Things will ssloowwww as the number of machines increases
What other choice do we have ?
I do feel that each task should iterate through the list of machines and despatch them all at once. If something bad happens we can just do a rollback.
We don't gain anything by waiting for one machine to finish all the tasks and then moving on to the other.
This feature has been created and tested for standard deploy. It will be merged to master when rolling_deploy branch is finished
Instead of having each machine iterate through the task list, it will have each task iterate through the machine list.