tobilg / docker-mongodb-marathon Goto Github PK

¡Hola!

I am trying to deploy a replica set on ECS (+route 53?); I am thinking about making a derivate of your image for that propouse. Would you kindly point me to some document about the node.js configuration app your'e using on it?

¡Gracias!

hi,
looking at this code,

     if (isInitialRun) {
          marathon.setupEventCallback(options.host, options.webPort.public)
         ...
         mongoDb.initializeReplicaSet()
                      .then(function(result) {
                        // Now we are ready to handle the events from Marathon
                        isReady = true;

unless I missed something, only the first instance that registers a node with zk can create a marathon event listener, and is also the only instance that isReady is ever == true.

if my first instance (in a replicaset) dies, and a new mongo master is elected, how can it's instance ever subscribe or react? (because even if it subscriibes, its isReady will always be false).

what am I missing?

Hi, nice work.
I'm running mongodb on marathon/mesos. Our main issue is that the process often dies as it exceeds any memory set in the marathon job. We have 32 Gb servers and allocate 4Gb to mongo. Usually we have 20Gb free when the 4Gb mongo is exceeded. Is there a way to un-restrict the memory for a task?

oversubscription seems to be a possible mechanism, placing mongo as the only priority task and letting all others run as over-subscribed. Have you tackled this? What about if mongo exceeds 32Gb, can we spill to swap somehow?
https://mesosphere.com/blog/2015/08/26/turbocharging-your-mesos-cluster-with-oversubscription/

Hmm. Looks like it will have to wait till marathon catches up
mesosphere/marathon#2424