Node Availability

• Each node can have one of three admin-controlled states
• Only affects if existing or new containers can run on that node
  • active: Runs existing tasks, available for new tasks
  • pause: Runs existing tasks, not available for new tasks
    • (good for troubleshooting)
  • drain: Reschedules existing tasks, not available for new tasks
    • (good for maintenance)
• This affects service updates and recovering tasks too
• Internet examples include "drain managers", but not realistic
  • Use labels to control manager tasks

Examples

• Prevent node2 from starting new containers
  • docker node update --availability pause node2
• Stop containers on node3 and assign their tasks to other nodes
  • docker node update --availability drain node3
• Mark node as available again
  • docker node update --availability=active
    • This does not rebalance containers across nodes

Hands On

[node1] ~> docker service create --name webapp1 --replicas 4 nginx
vcc4y9ootqso9y33ibc3p49p0
overall progress: 4 out of 4 tasks
1/4: running   [==================================================>]
2/4: running   [==================================================>]
3/4: running   [==================================================>]
4/4: running   [==================================================>]
verify: Service converged

[node1] ~> docker node update --availability=pause manager2
manager2
##### VISUALIZER IS EXACTLY SAME AS ABOVE

[node1] ~> docker service update --replicas=8 webapp1
webapp1
overall progress: 8 out of 8 tasks
1/8: running   [==================================================>]
2/8: running   [==================================================>]
3/8: running   [==================================================>]
4/8: running   [==================================================>]
5/8: running   [==================================================>]
6/8: running   [==================================================>]
7/8: running   [==================================================>]
8/8: running   [==================================================>]
verify: Service converged

[node1] ~> docker node update --availability=active manager2
manager2
##### VISUALIZER IS EXACTLY SAME AS ABOVE
##### SWARM by default DOES NOT rebalance cluster when nodes become active again

[node1] ~> docker node update --availability=drain worker1
worker1

[node1] ~> docker node ls
ID                            HOSTNAME            STATUS              AVAILABILITY        MANAGER STATUS      ENGINE VERSION
th2nb6uc5j61iqsz67q1oztaj *   manager1            Ready               Active              Leader              18.09.1
rru8z0gj9riq61jp3fwiyiesq     manager2            Ready               Active                                  18.09.1
tsojlv10aqlvncyr58ptg1q88     worker1             Ready               Drain                                   18.09.1

[node1] ~> docker node update --availability=active worker1
worker1
##### VISUALIZER IS EXACTLY SAME AS ABOVE
##### SWARM by default DOES NOT rebalance cluster when nodes become active again

External links

• https://docs.docker.com/engine/swarm/manage-nodes/\#change-node-availability