research: docker swarm

[ilyam8]

Feature idea summary

Docker Swarm is a clustering and scheduling tool for Docker containers.

Preparation to: #5710

Improvement of docker_engine collector.

Plan is:

• read documentation
• set up a lab
• find how to collect metrics
• see exposed metrics
• group metrics into charts
• understand what alarms we need

[ilyam8]

Ok, i didn't dig into it but at a glance docker in swarm mode is very nice.

In Swarm mode docker engine exposes following metrics (actually much more, but others don't make sense for me atm, i may be wrong here ofc):

# HELP swarm_manager_configs_total The number of configs in the cluster object store
# TYPE swarm_manager_configs_total gauge
swarm_manager_configs_total 0
# HELP swarm_manager_leader Indicates if this manager node is a leader
# TYPE swarm_manager_leader gauge
swarm_manager_leader 1
# HELP swarm_manager_networks_total The number of networks in the cluster object store
# TYPE swarm_manager_networks_total gauge
swarm_manager_networks_total 3
# HELP swarm_manager_nodes The number of nodes
# TYPE swarm_manager_nodes gauge
swarm_manager_nodes{state="disconnected"} 0
swarm_manager_nodes{state="down"} 0
swarm_manager_nodes{state="ready"} 3
swarm_manager_nodes{state="unknown"} 0
# HELP swarm_manager_secrets_total The number of secrets in the cluster object store
# TYPE swarm_manager_secrets_total gauge
swarm_manager_secrets_total 0
# HELP swarm_manager_services_total The number of services in the cluster object store
# TYPE swarm_manager_services_total gauge
swarm_manager_services_total 1
# HELP swarm_manager_tasks_total The number of tasks in the cluster object store
# TYPE swarm_manager_tasks_total gauge
swarm_manager_tasks_total{state="accepted"} 0
swarm_manager_tasks_total{state="assigned"} 0
swarm_manager_tasks_total{state="complete"} 0
swarm_manager_tasks_total{state="failed"} 0
swarm_manager_tasks_total{state="new"} 0
swarm_manager_tasks_total{state="orphaned"} 0
swarm_manager_tasks_total{state="pending"} 0
swarm_manager_tasks_total{state="preparing"} 0
swarm_manager_tasks_total{state="ready"} 0
swarm_manager_tasks_total{state="rejected"} 0
swarm_manager_tasks_total{state="remove"} 0
swarm_manager_tasks_total{state="running"} 5
swarm_manager_tasks_total{state="shutdown"} 0
swarm_manager_tasks_total{state="starting"} 0
# HELP swarm_node_info Information related to the swarm
# TYPE swarm_node_info gauge
swarm_node_info{node_id="193816ofdqsg9kkm0hkfladvo",swarm_id="k1a6iu49n97a1vej3u5pjgsbr"} 1
# HELP swarm_node_manager Whether this node is a manager or not
# TYPE swarm_node_manager gauge
swarm_node_manager 1

The key thing that only nodes with Manager role show real numbers, Worker nodes show zeroes (in current version, i think it will be fixed later).

So additional charts will be added only for Manager nodes.