Thank you for opening your first issue here! 🛠

An alternative could be using the Pod plugin, but any env specified there is overwritten and when trying to use env_from serialization fails.
Also ContainerTask currently doesn't allow specifying Pod via task_config...

Closing this as it can be now achieved with PodTemplates:
#3123
flyteorg/flytekit#1515

I'm currently facing this issue, and would really appreciate a name parameter to control a secret's name, for exactly the reasons listed above.

My use-case is that I have local packages that are (and should be) unaware of Flyte and Kubernetes; using environment variables that get read automatically with names like POSTGRES_PASSWORD. When injecting them in Flyte through Kubernetes Secrets, they become _FSEC_DB-CREDENTIALS_POSTGRES-PASSWORD, which I then have to manually change again, duplicating code (and environment variables) all over.

It seems very convoluted to do this via PodTemplate, as I don't care about configuring the rest of the Pod spec and it clutters the workflow (similar to the comment made in #3123 (comment)).

Please consider reopening @flixr .

NB: Similarly for Secrets mounted as files, would be great to be able to control the mountPath from the Task decorator.

@gpgn thank you for your he comments. I actually not sure
If everything is controllable at the k8s layer, but this is a fair ask.
Cc @flixr / @EngHabu / @eapolinario

Ive made a draft implementation here:

• Introduces new "env_var" and "file" fields to Secret to allow specifying name/mountPath on injection flytekit#1726
• Introduces new "env_var" and "file" fields to Secret to allow specifying name/mountPath on injection #minor flytepropeller#589
• Introduces new "env_var" and "file" fields to Secret to allow specifying name/mountPath on injection flyteidl#423
• Adds documentation on usage for new Secret fields env_var and file flytesnacks#1017

Quite new to Flyte (and not a Go dev), so probably I'm missing context and it needs more work in other places where Secrets are used, but my (small) local experiments work. Would be happy to improve it.

(also many thanks to @Yicheng-Lu-llll and @jeevb for help with setting up a local dev environment! 🙏 )

Updated the PRs to allow for specifying mountPath as well. Both old and new way work:

from flytekit import task, workflow, Secret
import flytekit
import os


@task(
    secret_requests=[
        Secret(
            group="user-info",
            key="user_secret",
            mount_requirement=Secret.MountType.FILE,
        ),
        Secret(
            group="user-info",
            key="user_secret",
            mount_requirement=Secret.MountType.ENV_VAR,
        ),
    ]
)
def old() -> None:
    context = flytekit.current_context()
    print("old")

    # mount_requirement=ENV_VAR
    print(context.secrets.get(env_name="_FSEC_USER-INFO_USER_SECRET"))

    # mount_requirement=FILE
    with open('/etc/flyte/secrets/user-info/user_secret', 'r') as infile:
        print(infile.readlines())
    return True

@task(
    secret_requests=[
        Secret(
            group="user-info",
            key="user_secret",
            env_var=Secret.MountEnvVar(name="foo"),
        ),
        Secret(
            group="user-info",
            key="user_secret",
            file=Secret.MountFile(path="/foo"),
        ),
    ]
)

def new() -> None:
    context = flytekit.current_context()
    print("new")

    # env_var=
    print(context.secrets.get(env_name="foo"))

    # file=
    with open('/foo/user_secret', 'r') as infile:
        print(infile.readlines())


@workflow
def training_workflow(hyperparameters: dict) -> None:
    """Put all of the steps together into a single workflow."""
    old()
    new()