It's a limitation: apparently your service is configured in a way that 2 cannot be running on the same host. Ideally, you'd fix the service (using SRV records is great for connecting services as they include a port), but if you only need one, and *do* have other nodes, AFAIK, you can do 2 things: - Enforce a unique host: https://mesosphere.github.io/marathon/docs/constraints.html - Add ports as resources ( http://mesos.apache.org/documentation/attributes-resources/), and define them in Marathon as a requirement (though admittedly I've never tried that myself, if that's what you're doing and it's not working, than it'd be bug IMHO).
On Tue, Nov 17, 2015 at 1:41 PM, Ahmet Emre Aladağ <[email protected]> wrote: > Hi, > > We are using Mesos 0.25 and Marathon 0.11.1 > > I don't know if this is fixed in the recent version of Marathon or 0.26-RC > version of Mesos. > > We have started an app on Marathon. The app uses service port 10000. > Manually entered this fixed port. > > When I try to restart the app, it tries to create a new task before > killing the existing one. It isn't able to acquire the resource: port 10000 > . So hangs on "Staged". > > I can't kill or suspend or restart or destroy it. It lives there forever > with 2 tasks (one Running, one Staged). Destroy responds with this message: > "Error destroying app: Futures timed out after [10000 milliseconds]" > > Interestingly it tried to start the new task on the same machine, not on > the other 2 slaves. > > Is this a bug? > > Thanks, >
