Depending on network reliability and speed, the AMP deployment may fail

[ndegory]

The timeout definition in the AMP stack files (cluster/agent/stacks/) is based on the time spent between the service creation/update request and all tasks in the expected state. One significant part of this time is spent pulling the Docker image (when it's not already there), which brings a high variability depending on the host where it's running, depending on the performance of Docker Hub and depending on what has already been pulled.
Raising the timeout to excessively high values to avoid a Error: context deadline exceeded is not a satisfying solution, there should be a check on the status of the tasks to identify when the image has been pulled, to start waiting for the timeout.

[ndegory]

PR #1628 seems to be broken because of this issue