Unacceptable slowness

[sheepdragon]

The following workflow executes a diamond workflow

# The following workflow executes a diamond workflow
# 
#   A
#  / \
# B   C
#  \ /
#   D
apiVersion: argoproj.io/v1alpha1
kind: Workflow
metadata:
  generateName: dag-diamond-
spec:
  entrypoint: diamond
  templates:
  - name: diamond
    dag:
      tasks:
      - name: A
        template: echo
        arguments:
          parameters: [{name: message, value: A}]
      - name: B
        depends: "A"
        template: echo
        arguments:
          parameters: [{name: message, value: B}]
      - name: C
        depends: "A"
        template: echo
        arguments:
          parameters: [{name: message, value: C}]
      - name: D
        depends: "B && C"
        template: echo
        arguments:
          parameters: [{name: message, value: D}]

  - name: echo
    inputs:
      parameters:
      - name: message
    container:
      image: alpine:3.7
      imagePullPolicy: IfNotPresent
      command: [echo, "{{inputs.parameters.message}}"]

In this example, no time-consuming operation is actually done. In theory, it should be finished very quickly. Why is the operation so slow and even slower after adding the actual logic

[sheepdragon]

Time is consumed in this place.

Why is it so slow？

[sarabala1979]

It looks like Pod creation and pod completions are taking time. Can you check K8s API server timing(POD events)?

[sheepdragon]

Is it reasonable to consume this time?
Whether this function is as slow as I use it under normal circumstances, I always doubt that I use it incorrectly.

[sheepdragon]

apiVersion: argoproj.io/v1alpha1
kind: Workflow
metadata:
  generateName: dag-diamond-coinflip-
spec:
  entrypoint: diamond
  templates:
  - name: diamond
    dag:
      tasks:
      - name: A
        template: echo
      - name: B
        depends: "A"
        template: echo

  - name: echo
    container:
      image: alpine:3.6
      command: [sh, -c]
      args: ["echo 123"]

This is a simple example. Both sides are executed, eliminating the time-consuming of image pulling. Is time spent on k8s scheduling? Is it my k8s configuration?

[sarabala1979]

Based on above screen shot. Pod Creation and schedule is taking 8s. Can you try it in your load k3d/docker on deshtop?

[sheepdragon]

pns.go

It looks like a forced wait

[sarabala1979]

Wait container will wait for the main container to complete. if you think PNS executor has a problem, can you try docker/k8sapi executor?

[sheepdragon]

I've verified that docker / k8sapi consumes the same amount of time.

Because I just used Argo for a short time, I used drone as a CI tool before. Drone does not have this problem. I want to eliminate the problem of Argo environment configuration. I want to know whether the above example consumes the same time running on your environment.