What is the usage of ARGO_TEMPLATE? Can we delete it?

No.

However, we could replace it with just the information needed by the emissary. The emissary only needs to know which containers it must wait on. It does not need the whole template.

That would be nice. Look forward to the coming improve!

The emissary only needs to know which containers it must wait on.

@alexec Looks like tests are breaking after I removed the parameters. #9698 Any hints on what I may have missed?

At Intuit we are actually relying on the $ARGO_TEMPLATE environment variable inputs because we had an issue in which we were originally trying to do this in our Step:

image: "docker.intuit.com/dev/patterns/kubernetes/dev/kubectl-awscli:v1.22.15"
command:
  - sh
  - -c
args:
  - |
    echo "{{`{{inputs.parameters.manifest}}`}}" | ....

but we found that we exceeded the maximum argument size limit allowed by Linux. So, we changed the Step to use the ARGO_TEMPLATE environment variable:

image: "docker.intuit.com/dev/patterns/kubernetes/dev/kubectl-awscli:v1.22.15"
command:
  - sh
  - -c
args: ["
  getconf ARG_MAX;
  echo $ARGO_TEMPLATE | yq '.inputs.parameters[0].value' --unwrapScalar > /tmp/manifest-64;
  ....
"]

(We had also tried an alternative solution of mounting the Input parameter as a RawArtifact but that faced an issue and didn't work either)

now perhaps there could be an alternative solution in which the value of the environment variable is compressed or something...

but we found that we exceeded the maximum argument size limit allowed by Linux. So, we changed the Step to use the ARGO_TEMPLATE environment variable:

Did this work for you? I'm having exact same issue.