API "/api/compute_resources/:compute_resource_id/available_clusters/:cluster_id/available_networks returning error

rahul · March 10, 2021, 12:04pm

Problem: I am trying to get available networks of particular compute resource with API "API “/api/compute_resources/1/available_clusters/domain-s7/available_networks” in Foreman v2.1 but it is throwing error “Internal Server Error: the server was unable to finish the request. This may be caused by unavailability of some required service, incorrect API call or a server-side bug. There may be more information in the server’s log”. I had checked log and it is showing "Backtrace for ‘Action failed’ error (NoMethodError): undefined method network for nil:NilClass"

Expected outcome: List of available networks should be returned

Foreman and Proxy versions: 2.1 & 2.1.0-rc3

Foreman and Proxy plugin versions: 2.1 & 2.1.0-rc3

Distribution and version: CentOS v7.x86_64 & v2.1

ofedoren · March 10, 2021, 12:11pm

@shiramax, @yifatmakias you might want to take a look at this.

rahul · March 10, 2021, 12:27pm

The compute resource in this case is VMware provider.

tbrisker · March 10, 2021, 12:52pm

can you please share the full stack trace from /var/log/production.log?

yifatmakias · March 10, 2021, 1:29pm

From what I checked so far it looks like this issue happens only in VMware compute resource. I will continue checking this.

rahul · March 11, 2021, 4:51am

foreman-log.log (9.0 KB)

rahul · March 11, 2021, 9:13am

Yes, I agree it is happening for VMware compute resource.

evgeni · March 11, 2021, 11:46am

For VMware, you need to provide the cluster name, not id for those calls.

See also Bug #31874: API uses cluster_id inconsistently - Foreman

rahul · March 12, 2021, 12:01pm

Thankyou, Your suggestion works for me.

evgeni · March 12, 2021, 1:03pm

Cool.

@yifatmakias @shiramax is there a way we could make the API accept both? Like it does for oVirt?

ezr-ondrej · March 13, 2021, 3:15pm

Not easily unfortunately, vmware itself don’t understand the concept of ID much and it boils down to this method:

github.com

vmware/rbvmomi/blob/dc3febf3cd49d85df598981676f9670c1bfa1733/lib/rbvmomi/vim/Datacenter.rb#L7


# frozen_string_literal: true
# Copyright (c) 2011-2017 VMware, Inc.  All Rights Reserved.
# SPDX-License-Identifier: MIT
class RbVmomi::VIM::Datacenter
  # Traverse the given inventory +path+ to find a ComputeResource.
  def find_compute_resource path
    hostFolder.traverse path, RbVmomi::VIM::ComputeResource
  end
  # Find the Datastore with the given +name+.
  def find_datastore name
    datastore.find { |x| x.name == name }
  end
  # Traverse the given inventory +path+ to find a VirtualMachine.
  def find_vm path

We probably could get all clusters and just .detect the one with appropriate ID, but we I’d rather if we move away from using IDs in vmware and just use paths as ids. But I’m really unsure what vmware prefers, that is just my guess from their api documentation and its ruby bindings.

evgeni · March 13, 2021, 3:40pm

But for other things (networks, datastores) it does want the ID when I deploy a VM.

What’s a path? Is that when clusters are nested? How does it look like?

ezr-ondrej · March 13, 2021, 9:46pm

you can nest all the objects in vmware into folders, so it coul look like RootFolderName/nestedfoler_name/cluster_name

We’ve done some nasty stuff for networks, but that’s broken for a lot of corner cases. I don’t think datastore takes ID though.