Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Hyphens in names cause the service orchestrator to start a grpc server #1850

Closed
mattpiatkus opened this issue May 19, 2020 · 3 comments
Closed
Projects
Milestone

Comments

@mattpiatkus
Copy link

mattpiatkus commented May 19, 2020

Seldon version 1.1
Kubernetes version 1.14

On my cluster, the following starts a grpc orchestrator rather than a rest one:

apiVersion: machinelearning.seldon.io/v1
kind: SeldonDeployment
metadata:
  name: fixed
spec:
  name: fixed-hyphenated
  protocol: seldon
  transport: rest
  predictors:
  - componentSpecs:
    - spec:
        containers:
        - image: seldonio/fixed-model:0.1
          name: classifier1
    graph:
      children: []
      name: classifier1
      type: MODEL
    name: default
    replicas: 1

With fixed-hyphenated reverted to just fixed, it correctly starts a rest server.

The problem also seems to occur with hyphens in the repository name, predictor name, etc. Repository names are a particular problem because (in my case) the name comes from the Google cloud project which can't be changed. I was only able to use a rest orchestrator for my image by hosting it on an entirely new project with no hyphens in the name.

@adriangonz adriangonz added the triage Needs to be triaged and prioritised accordingly label Jun 2, 2020
@ukclivecox
Copy link
Contributor

I tried this on 1.1.0 and seems to work fine. Can you check the output of kubectl get sdep -o yaml
I get:

apiVersion: v1
items:
- apiVersion: machinelearning.seldon.io/v1
  kind: SeldonDeployment
  metadata:
    creationTimestamp: "2020-06-02T13:47:23Z"
    generation: 1
    name: fixed
    namespace: default
    resourceVersion: "2147"
    selfLink: /apis/machinelearning.seldon.io/v1/namespaces/default/seldondeployments/fixed
    uid: 66a89de8-78ac-4648-9b78-c9151adf67f8
  spec:
    name: fixed-hyphenated
    predictors:
    - componentSpecs:
      - metadata:
          creationTimestamp: null
        spec:
          containers:
          - image: seldonio/fixed-model:0.1
            name: classifier1
            ports:
            - containerPort: 6000
              name: metrics
              protocol: TCP
            resources: {}
            volumeMounts:
            - mountPath: /etc/podinfo
              name: podinfo
      engineResources: {}
      graph:
        endpoint:
          service_host: localhost
          service_port: 9000
          type: REST
        implementation: UNKNOWN_IMPLEMENTATION
        name: classifier1
        type: MODEL
      labels:
        version: default
      name: default
      replicas: 1
      svcOrchSpec: {}
    protocol: seldon
    transport: rest
  status:
    address:
      url: http://fixed-default.default.svc.cluster.local:8000/api/v1.0/predictions
    deploymentStatus:
      fixed-default-0-classifier1:
        availableReplicas: 1
        replicas: 1
    replicas: 1
    serviceStatus:
      fixed-default:
        grpcEndpoint: fixed-default.default:5001
        httpEndpoint: fixed-default.default:8000
        svcName: fixed-default
      fixed-default-classifier1:
        httpEndpoint: fixed-default-classifier1.default:9000
        svcName: fixed-default-classifier1
    state: Available
kind: List
metadata:
  resourceVersion: ""
  selfLink: ""

I install Ambassador and port-fowarded to 8003 and ran:

curl -d '{"data": {"ndarray":[[1.0, 2.0, 5.0]]}}'    -X POST http://localhost:8003/seldon/default/fixed/api/v1.0/predictions    -H "Content-Type: application/json"

{"data":{"names":[],"ndarray":[1,2,3,4]},"meta":{}}

@ukclivecox ukclivecox removed the triage Needs to be triaged and prioritised accordingly label Jun 4, 2020
@axsaucedo axsaucedo added this to To do in 1.3 via automation Jul 15, 2020
@axsaucedo axsaucedo added this to the 1.3 milestone Jul 15, 2020
@axsaucedo
Copy link
Contributor

@mattpiatkus any updates?

@ukclivecox
Copy link
Contributor

Please reopen if can be reproduced

1.3 automation moved this from To do to Done Aug 20, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
No open projects
1.3
  
Done
Development

No branches or pull requests

4 participants