Noticing that with a deployment it's fairly common for an application to get "stuck" running above capacity. This is kind of a drag, but especially frustrating because it leaves the app in a perpetual "Deploying" state.
I'm also seeing messages that follow this pattern in the logs:
Jun 21 18:01:13 ip-xx-xx-xx-xx marathon: [2016-06-21 18:01:12,962] INFO Task launch for 'task [xxx.23fe386f-37da-11e6-b604-024258f8ee42]' was accepted. 0 tasksToLaunch, 0 in flight, 516 confirmed. instance count delta 16. not backing off (mesosphere.marathon.core.launchqueue.impl.AppTaskLauncherActor:marathon-akka.actor.default-dispatcher-249)
I haven't been able to detect a clear pattern in the applications that are impacted, some are running many instances 500 - 1000, others are running a few 30-50.
Marathon version: 1.1.1
Mesos version: 0.28.2
The only configuration applied to the Marathon process is --task_launch_timeout 300000
Happy to provide additional details around our setup or surface extra information from the logs if that would be helpful.