Faas: Question on limits for a Swarm cluster

[pawelmarkowski]

Configuration attached below does not be reflected in

docker service inspect hello

provider:
  name: faas
  gateway: http://127.0.0.1:8080
functions:
  hello:
    lang: python3
    handler: ./hello
    image: hello:latest
    labels:
        "com.openfaas.scale.min": "2"
        # "com.openfaas.scale.max": "40"
    limits:
      cpu: 1000m
      # memory: 1000m
    requests:
      cpu: 1000m
      # memory: 400m

currently inspect returns "Resources": {}, I will paste below results from the code with uncommented memory params.

limits:
      cpu: 1000m
      memory: 1000m
requests:
      cpu: 1000m
      memory: 400m

Expected Behaviour

Reservations for CPU in a service parameters

Current Behaviour

"Resources": {
"Limits": {
"MemoryBytes": 1048576000
},
"Reservations": {
"MemoryBytes": 419430400
}
},

Steps to Reproduce (for bugs)

1. faas-cli new --lang python3 hello
2. update hello.yml file
3. faas-cli build -f hello.yml
4. faas-cli deploy -f hello.yml

Your Environment

CLI:
commit: a141dedf94ffeed84412365fd591bdc8999c5a1b
version: 0.8.3

Gateway
uri: http://127.0.0.1:8080
version: 0.9.11
sha: 78a4584019a2a5d86b606a00b4fba979930cc2a7
commit: Add sha256 hashgen for watchdog to ci config

Provider
name: faas-swarm
orchestration: swarm
version: 0.5.0
sha: 4269ab15673ad883bfdce4df47c055588e8e86b2

• Docker version docker version Version: 18.09.0

• Operating System and version (e.g. Linux, Windows, MacOS): Fedora

[pawelmarkowski]

IMO it is caused by different notation in CPU limits / requests in a docker swarm
it is not cpu: 1000m but cpus: 0.5

[alexellis]

Derek add label: support

[alexellis]

Derek add label: question

[alexellis]

https://github.com/openfaas/faas/issues/1026#issuecomment-453748305

Sounds like it might be right. Where did you get the values from that you used?

[pawelmarkowski]

I used config described in the docs https://docs.openfaas.com/reference/yaml/#function-memorycpu-limits

I checked version with:

provider:
  name: faas
  gateway: http://127.0.0.1:8080
functions:
  hello:
    lang: python3
    handler: ./hello
    image: hello:latest
    readonly_root_filesystem: true
    environment:
      read_timeout: 3600s
      write_timeout: 3600s
    labels:
        "com.openfaas.scale.min": "2"
        # "com.openfaas.scale.max": "40"
    limits:
      cpus: 1
      memory: 1000m
    requests:
      cpus: 0.5
      memory: 400m

and the result in my local swarm is the same

"Resources": {
"Limits": {
"MemoryBytes": 1048576000
},
"Reservations": {
"MemoryBytes": 419430400
}
},

what is proper way to limit cpus for the function if I use OpenFaaS and swarm? I was trying different combinations :)

[alexellis]

There are links on the page for both systems so that you can pick the correct option.

Here's the link from through page for Swarm https://docs.docker.com/config/containers/resource_constraints/

Alex

[pawelmarkowski]

Yes I know I wanted to add it in subcomment ;) https://docs.docker.com/config/containers/resource_constraints/ - but even if I use cpus option it does not work in OpenFaaS as you can see in my previous comment. (I have check cpu-period and so on also ...)

[alexellis]

The code and docs say "cpu" is the name of the field but your example says "cpus". I'm guessing that may be the problem?

https://github.com/openfaas/faas-cli/blob/master/stack/schema.go

Alex

[pawelmarkowski]

Yes in L61, I used cpu also as you can see in my first post. So it sounds like a bug in OpenFaaS - do you agree?

[alexellis]

The example you are using is wrong and can't work:

provider:
  name: faas
  gateway: http://127.0.0.1:8080
functions:
  hello:
    lang: python3
    handler: ./hello
    image: hello:latest
    readonly_root_filesystem: true
    environment:
      read_timeout: 3600s
      write_timeout: 3600s
    labels:
        "com.openfaas.scale.min": "2"
        # "com.openfaas.scale.max": "40"
    limits:
      cpus: 1
      memory: 1000m
    requests:
      cpus: 0.5
      memory: 400m

If there were an issue it would be with the openfaas/faas-swarm repo.

Have you looked at the logs of faas-swarm? docker service logs func_faas-swarm

Alex

[alexellis]

The documentation doesn't explain whether to use CPUs / CPU quota or CPU period in the field for Swarm. It just mentions Kubernetes.

From looking at the code I'd say it's CPU quota

https://github.com/openfaas/faas-swarm/blob/2dfeaf3933f4ebe0133642b9e087fe981660172f/handlers/resources_test.go#L127

From the Swarm docs:

Specify how much of the available CPU resources a container can use. For instance, if the host machine has two CPUs and you set --cpus="1.5", the container is guaranteed at most one and a half of the CPUs. This is the equivalent of setting --cpu-period="100000" and --cpu-quota="150000". Available in Docker 1.13 and higher.

[pawelmarkowski]

@alexellis yes you are right it is a problem with faas-swarm

func_faas-swarm.1.[email protected] | 2019/01/12 21:09:18 Error parsing cpu limit: *strconv.NumError
func_faas-swarm.1.[email protected] | 2019/01/12 21:09:18 Error parsing cpu reservations: *strconv.NumError

limits:
      cpu: 1.1

[alexellis]

That's not a problem with faas-swarm, you've just specified the limits wrong. See above.

[pawelmarkowski]

@alexellis sorry for my hasty answer. I have not seen your response :)
yes your assumtion based on the code was correct.

limits:
      cpu: 1000000

causes in docker service inspect hello

"Resources": {
                    "Limits": {
                        "NanoCPUs": 1000000,
                        "MemoryBytes": 1048576000
                    },
}

It would be great if somebody can mention about it in a docs as like memory case.

[pawelmarkowski]

thanks a lot @alexellis we can close the ticket. I will create another one, but it will be question / feature request ;)