Application upgrades with zero downtime

In certain cases it is desirable to upgrade applications and have zero, or at least minimize, application downtime.

To achieve zero downtime we require multiple replicas and strongly recommend to have multiple hosts.

Rolling upgrades

If your applications has multiple replicas, you can use simple rolling upgrades, where one service instance at a time will be upgraded.

demo app spec

name: my-application
version: "8"
services:
  - name: svc
    mode: replicated
    replicas: 3
    share-pid-namespace: false
    containers:
      - name: alpine
        container-log-size: 100 MB
        container-log-archive: false
        image: registry-1.docker.io/alpine
        cmd:
          - sh
          - -c
          - sleep infinite
        on-mounted-file-change:
          restart: true
on-mutable-variable-change: restart-service-instance
upgrade-from:
  - version-regexp: .*
    method: per-service
    services:
      - name: svc
        instances-in-parallel: 1
        healthy-time: 1m

The upgrade-from tells the system how many services can be upgraded in parallel and the healthy time (of the new containers) before upgrading the next one.