docker stats in Swarm mode?

Question

Im trying to setup docker in swarm mode and monitor the resource utilization of all the services/containers running in the swarm.

Docker stats on the manager node doesnt seem to show the resource utilization on the worker nodes.

Is there any way I can do this?

Thanks.

Im looking for something like docker stats, through which I can get the utilization easily, like using the api, does grafana expose some api to do this? — grattttt, Aug 27 '17 at 18:36

score 4 · Answer 1 · edited Apr 21 '20 at 20:39

4

Try Ansible:

ansible docker -a "docker stats --no-stream"

Where you setup your "docker" nodes in /etc/ansible/hosts

edited Apr 21 '20 at 20:39

David Buck

3,752
35
31
35

answered Apr 21 '20 at 17:26

Andrew G

51
2

exactly the answer the OP was (rightly) looking for; all it requires is ansible on the manager + automated way to log into other docker hosts from manager, which is likely the case anyway; produces `docker stats` output for each of the docker hosts. Cool. Thx. – P Marecki Jun 15 '20 at 11:21

score 2 · Accepted Answer · answered Aug 27 '17 at 19:27

There's no direct way to retrieve all container stats of a given service in a Swarm. You'll probably have to use more steps to discover all tasks of a service, all node addresses, and each container id. The engine api docs should help you getting started. If you need some inspiration, I'd suggest you to peek into such overview dashboards like the https://github.com/charypar/swarm-dashboard or the https://github.com/dockersamples/docker-swarm-visualizer.

score 0 · Answer 3 · answered Dec 24 '19 at 07:37

0

Try this for CPU and Memory usage

docker stats --all --format "table {{.ID}}\t{{.Name}}\t{{.CPUPerc}}\t{{.MemUsage}}"

answered Dec 24 '19 at 07:37

Nikita Shesterikov

359
2
8

1

displays stats only for the host, where run; no swarm info – P Marecki Jun 15 '20 at 11:01

score 0 · Answer 4 · answered Jul 22 '21 at 06:27

You can use cadvisor to deploy on each node or a swarm service extended on all nodes and collect metrics one by node.

docker run -d --name=cadvisor -p 8080:8080 --volume=/var/run:/var/run:rw --volume=/sys:/sys:ro --volume=/var/lib/docker/:/var/lib/docker:ro google/cadvisor:latest

user548649 · Answer 5 · 2022-01-13T11:19:04.530

You can do this for all worker nodes using the following bash command:

docker node ls | grep -v Leader | grep -v Reachable | cut -c 31-47 | grep -v HOSTNAME | xargs -I"SERVER" sh -c "echo SERVER; ssh SERVER docker stats --no-stream"

This command works as follows.

List all Docker swarm nodes:

docker node ls

Filter out the manager nodes:

grep -v Leader | grep -v Reachable

Select the ip-addresses of the worker nodes:

cut -c 31-47

Remove the column header from the result:

grep -v HOSTNAME

Print the ip-address of the worker node and execute docker stats on the worker node:

xargs -I"SERVER" sh -c "echo SERVER; ssh SERVER docker stats --no-stream"

docker stats in Swarm mode?

5 Answers5