Skip to content

Latest commit

 

History

History
116 lines (85 loc) · 5.35 KB

CHANGELOG.md

File metadata and controls

116 lines (85 loc) · 5.35 KB
Raw

Change Log

Current status

  • Vagrant spins up 6 virtual box machines, 3 managers for quorum & 3 workers

  • Ansible provisions the boxes installing docker 1.12

  • The primary manager starts the swarm cluster, each other manager and worker then joins

    • The status of the swarm is outputted at the end

  • Installation of docker & initialisation of the swarm is idempotent so provision can be re-run

    • Use of shell can be replaced when there are Ansible modules for managing swarm.

  • Start collection of services to test the swarm

  • multiple overlay networks for front and backend

  • Clustered consul servers and nodes on all hosts

  • Centralised logging via syslog docker log driver feeding into ELK stack

    • Will publish all the endpoints

  • Host monitoring using collectd -> riemann -> influxdb -> grafana.

Roadmap

  • Add monitoring, investigate

    • cAdvisor for docker stats
    • collectd
    • Riemann tools for Docker & Consul
    • statsd for udp from applications (though Riemann.io recommend against UDP)

  • Add shipyard for visualisation

  • Consider Alpine vagrant box for workers as only running Docker

  • Generate ca & certs so we can connect to the manager1 daemon from other boxes.

  • Update systemd docker file on manager1 so we can start the daemon with TCP/TLS support

  • Look to start daemon running on 2375 https://docs.docker.com/v1.10/engine/reference/commandline/daemon/

  • Add multiple overlay networks and pull out hardcoded appnet configuration

  • Correct consul-notifier to handle multiple ports and register the name with the port in the same way registrator does.

0.5 (Unreleased)

Added:

0.4 (2016-08-26) Monitoring

Added:

  • Added monitoring ansible playbook
  • Added global collectd container service
  • Added influxdb service constrained to influx ansible group
  • Added single instance of Riemann. H/A is complicated
  • Added grafana service including influxdb datasource and demo dashboard for CPU - needs work

0.3 (2016-08-20) Logging and Consul

Added:

  • Created python script that utilises docker-py and python-consul to listen for docker start/stop events and register/de-register with consul. The script runs inside an Alpine python container on each node in the cluster.

    By default the container listens to the stream. You can manually register/deregister by passing args to the script

    /app # [root@worker1 vagrant]# docker exec -it $(cname consul-notifier) /bin/ash
/app # python consul-notifier.py -a register -n consul-notifier
[2016-08-18 12:18:13,374] Consul notifier processing register
[2016-08-18 12:18:13,377] Registering consul-notifier worker1:consul-notifier:80 port 80
/app # python consul-notifier.py -a deregister -n consul-notifier
[2016-08-18 12:18:18,184] Consul notifier processing deregister
[2016-08-18 12:18:18,188] De-registering consul-notifier worker1:consul-notifier:80

  • Added logging.yml and monitoring.yml and Vagant ansible invocations.

  • Added rsyslog Ansible for all boxes

  • Added syslog driver and tags to consul-notifier

  • Added syslog to all services

  • Added rudimentary ELK stack on 2nd Manager box

    • rsyslog runs on every box and forwards messages to logstash -> elasticsearch -> kibana on Manager2

Issues:

  • Registrator does not work with 1.12 and service events Address above

    Possible temporary solution is to have a container on each node and poll the docker daemon for events

    echo -e "GET /events?since=1471083135 HTTP/1.0\r\n" | nc -U /var/run/docker.sock

    The docker API returns newline delimited JSON which can be parsed using NewlineJson

    This information can then be used to update consul for stopped and started apps

    /v1/agent/service/deregister/<serviceId>
/v1/agent/service/register/<serviceId>

0.2 (2016-08-13)

Added:

  • Restrict services to worker nodes using constraint filter
  • Added clustered consul docker container to each machine
    • All provisioned with Ansible
    • 3 servers on each manager
    • 3 nodes on each worker
    • UI available at http://192.168.77.21:8500/ui
    • DNS available at dig @172.17.0.1 -p 53 manager2.node.consul

0.1 (2016-08-11)

Added:

  • Added multi-provisioner, multi-box Vagrant file to spin up Docker swarm managers and workers
  • Added Ansible provisioning swarm.yml to bootstap managers and workers
  • Added apps.yml playbook to start up a collection of demo containers to demonstrate swarm

Known Issues

  • docker service rm seems to leave containers in /dev/mapper causing very high i/o wait