etcd-snapshot loading config fails with "flag provided but not defined: -profile"

[mitchellmaler]

Environmental Info:
RKE2 Version:
rke2 version v1.21.6+rke2r1 (b915fc9)
go version go1.16.6b7

Node(s) CPU architecture, OS, and Version:
N/A

Cluster Configuration:
N/A

Describe the bug:
Running the etcd-snapshot commands throw an error flag provided but not defined: -profile.

Steps To Reproduce:
Have the profile set in the config file and run the etcd-snapshot which loads the default config.

Expected behavior:
Runs snapshot.

Actual behavior:
Throw error saying flag profile is not defined.

[brandond]

cc @dereknola

etcd-snapshot now loads the config file, but if there are other flags in the config file that aren't valid for the etcd-snapshot command it will barf on them. This same problem prevents people from using the same config file for both server and agent, as the agent will complain about unknown server parameters. I wonder if we should just ignore unknown flags?

[Oats87]

Ignoring unknown flags can be dangerous, as our current behavior of barfing on unknown flags allows for a user to get more immediate feedback that they have a misconfiguration. If we just ignore unknown flags a user can be misled especially as that is a departure from existing behavior.

[pitshou]

I have a customer who is hitting this. So, I am checking to see what they have inside of their config file

Incorrect Usage. flag provided but not defined: -node-label

[dnoland1]

Assume you can pass the --config flag and choose an alternative config.yaml to workaround this, right?

[mitchellmaler]

yes there are ways around this. Different config or just passing the args/flags directly to the command.

[mitchellmaler]

The expected result though is the server config with all the server flags can be used with the etcd-snapshot command.

[brandond]

I agree that ignoring flags could cause other issues @Oats87 but the current behavior makes the config file non-portable between commands. If you add flags specific to rke2 server those will cause rke2 agent or rke2 etcd-snapshot (and planned future subcommands) to fail. This probably needs a proper design discussion to hash out since there's no easy fix that will satisfy all concerns.

[dereknola]

/backport v1.21.7+rke2r1

[dereknola]

/backport v1.21.7+rke2r1

[dereknola]

Backport bot broken?

[chreniuc]

I've encountered the same problem when trying to generate an on-demand snapshot. I'm using: rke2 version v1.21.6+rke2r1 (b915fc986e84582458af7131fe7f4e686f2af493).

The error shows up even when I run help:

rke2 etcd-snapshot help

FATA[0000] flag provided but not defined: -cluster-cidr 

The config file for the server contains:

cluster-cidr: 10.42.0.0/16

[brandond]

yes @chreniuc that's the same issue that will be resolved.

[VestigeJ]

Validated a fix across all RKE2 Release Candidates

#2130 [rke2 version v1.20.13-rc1+rke2r1]
#2129 [rke2 version v1.21.7-rc2+rke2r1]
#2191 [rke2 version v1.22.4-rc1+rke2r1]