mediative / mesos-stack Goto Github PK

Some of the dcos cli commands seems deprecated as the version installed is 2.0. e.g.
I could not get to install cassandra because I couldn't run

dcos package repo add Universe https://universe.mesosphere.com/repo

Any clues?

Hi, I am looking at something similar to run on baremetal and I am testing now with vagrant. Just a question of curiosity, why is 3.19 needed, is it because vivid uses systemd instead of upstart?

Motivation

Improve monitoring and recovery.

Input

• Installing Chronos as DCOS package: https://docs.mesosphere.com/services/chronos/
• DCOS Chronos package files: https://github.com/mesosphere/universe/tree/version-1.x/repo/packages/C/chronos/2
• Apollo DCOS role: https://github.com/Capgemini/Apollo/tree/master/roles/dcos_cli

Output

• Chronos role updated to run via Marathon
  -- Installed via DCOS or some a custom docker image
• Chronos UI is accessible on edge-node using HTTP basic or internally via chronos.marathon.mesos

Test

• Chronos is shown in Marathon UI
• dig chronos.marathon.mesos should resolve to a node IP
• Chronos UI is accessible on edge-node or via SSH port forwarding

Motivation

Better visibility on failures.

Input

• GCE tutorial: http://mesosphere.github.io/mesos-dns/docs/tutorial-gce.html
• Apollo DCOS role: https://github.com/Capgemini/Apollo/tree/master/roles/dcos_cli

Output

• Updated Mesos DNS role that
  -- Generates marathon template with proper hostname constraints and instances
  -- Deploys or updates the marathon template

Test

• Mesos DNS is shown in Marathon UI
• SSHing into any node and running dig master.mesos should show IP of a master node

Motivation

Marathon LB has been released and is available via DCOS

Input

• Marathon LB releases: https://github.com/mesosphere/marathon-lb/releases
• Apollo DCOS role: https://github.com/Capgemini/Apollo/tree/master/roles/dcos_cli

Output

• Use official release instead of snapshot
• Stretch: install via DCOS, or run via Marathon instead of installing as a node system service.

Test

• nginx.mesos.dev still works in the Vagrant cluster after the update

I got this when when I ran the playbook....

RUNNING HANDLER [mesos-master : Restart master services] ***********************
changed: [10.141.141.10] => (item=marathon)
changed: [10.141.141.10] => (item=chronos)
changed: [10.141.141.10] => (item=mesos-master)
failed: 10.141.141.10 => {"failed": true, "item": "mesos-dns", "msg": "no service or tool found for: mesos-dns"}

Motivation

Fix dcos node ssh --master and dcos node ssh --slave=X to work.

Input

dcos node ssh command does not work since it requires that the Mesos DNS API is exposed on one of the externally accessible nodes in the cluster under /mesos_dns.

$ dcos node ssh --master
Error while fetching [http://master.mesos:5050/mesos_dns/v1/hosts/leader.mesos.]: HTTP 404: Not Found

• Mesosphere's nginx router scripts: https://github.com/mesosphere/adminrouter-public

Output

• Mesos DNS API exposed on one of the externally accessible node(s) under /mesos_dns.

Motivation

Provide a way to persist application logs etc to S3

Input

• https://mesosphere.com/blog/2015/10/08/with-help-from-emc-were-bringing-enterprise-storage-to-mesos/

Output

• An ansible role which sets up an agent to persist mounted volumes to S3

Test/QA

• TBD

Motivation

The default Posix isolator doesn’t provide the same robust isolation guarantees that the cgroups isolator does.

Input

• Mesos agent configuration: http://mesos.apache.org/documentation/latest/configuration/

Output

• Enable cgroup isolation on the Mesos agent by passing the command line option --isolation='cgroups/cpu,cgroups/mem'

Test

• ???

Motivation

Keep the github project site in sync with the master branch.

Input

• https://www.npmjs.com/package/gh-pages

Output

• Travis CI script updated to build gitbook pages and push to gh-pages branch.

Test

• Last commit in gh-pages branch was done by Travis.