Some guidance required for building mesosphere solution on Amazon EC2 servers.
For the staging state of my application I was thinking of launching a master and a slave on a single instance plus another slave on a different instance. Plus marathon and chronos frameworks on top of mesos master node.
Marathon should be responsible for running http server on the same instance, which should serve as an access layer to the mesosphere API's.
Chronos will be used to run 1000s of very simple bash script (execution time about 1-5sec) each. (in this part not sure if aggregation of jobs is needed)
Should I run chronos via marathon, or just launch it as the service on my master node? Is there any difference?
Any guidance/advice on the current architecture design?
Is there a official support from Amazon to run mesosphere in amazon cloud?
You can run a single master mesos cluster only on two hosts but it will not be HA so there is no need to run two instances of anything except the mesos-slave.
If you really only want to use two hosts and use mesos I would recommend the following setup:
Host1:
Host2:
You need to configure the quorum number to 1 and run the cluster like that (no master election can happend that way, if Host1 dies, all processes stop executing). In practice you could run two instances of zookeeper and mesos-master (I saw a working setup like that) but in theory the leader election might screw up things because of the quorum setup of only 2 zookeeper nodes. My advice is to use at least 3 host (each host running a zookeeper, mesos-master, mesos-slave, marathon and chronos, new host's only need to run mesos-slave). This way you can manage with a host down.
If you really want to stick to the two host setup I already gave my advice, ask if I haven't answered something.
Also I don't know any official amazon support for mesos.