How to retrieve some pod ip ?

[EricBoix]

Consider the following workflow

# Hera version 4.4.2
from hera import (Task, Workflow)

with Workflow("retrieve-deamon-pod-ip-", generate_name=True) as w:

  daemon_t = Task(
        "startpostgresdb",
        daemon=True,
        image="postgres:15.2",
  )
  consumer_t = Task(
        "isdbready",
        image="postgres:15.2",
        command=["/bin/bash", "-c"],
        # Following line will trigger an "Internal Server Error" on submission
        # with error message
        #          templates.isdbready: failed to resolve {{tasks.startpostgresdb.ip}}
        args=["exec pg_isready -U user -h " + daemon_t.ip + " -p 5432"],
  )
  daemon_t >> consumer_t 

w.create()

will trigger on submission the error templates.isdbready: failed to resolve {{tasks.startpostgresdb.ip}}.
What am I doing wrong ?

Note: when using AW the syntax {{steps.startpostgresdb.ip}} (that is steps... as opposed to tasks...) was the one required for the workflow to be valid.

[flaviuvadan]

Thanks for the question @EricBoix! Any chance you looked at the daemon example? It has a sample usage of the task IP exposed via an env variable!

[EricBoix]

Thanks @flaviuvadan. Yes, I did encounter the deamon example that I could run smoothly while using the <some-task-name>.ip syntax.

Indeed for the example I took in this discussion, flowing the <some-task-name>.ip through the environment (the way deamon example does it) would work. But this is only because the pg_isready SQL related command accepts the environment variables supported by libpq (which includes PGHOST).

But some of my use cases (arising from yaml AW e.g. this workflow) are a bit different. They do require that<some-task-name>.ip be given/transmitted as inputs or arguments as opposed to environment variables.

Is this possible in 4.4.2 or later ?

[flaviuvadan]

@EricBoix yes! I just tested that. I can also put up an example with this if that's helpful. Here's the example:

from hera.shared import global_config
from hera.workflows import script, Workflow, Parameter, Steps

global_config.host = "..."
global_config.token = ..."
global_config.namespace = "..."

@script(daemon=True)
def daemon() -> None:
    import time 
    while True:  # this needs to stay "alive" for otherwise the IP is not exposed
        time.sleep(10)


@script()
def consume(ip: str) -> None:
    print(ip)


with Workflow(generate_name="daemon-ip-", entrypoint="s") as w:
    with Steps(name="s"):
        d = daemon()
        consume(arguments=Parameter(name="ip", value=d.ip))

w.create()

The above ran a workflow with 2 steps:

• first one stays alive because of the while True loop, which basically emulates a server that stays alive and scans for incoming requests
• second one prints the IP of the server that runs in the daemon step

[EricBoix]

Thanks @flaviuvadan . That nails it. I just added an input parameter to the Task/container
requiring the ip number of the server it consumes and then used it inside the args field.
Nice.
Closing the discussion.

[flaviuvadan]

Very glad to hear it's working for you in that setup @EricBoix!