Host creation is not idempotent

[tinova]

Description

Registering a host first time works as expected. Registering a second time removes it from the host cluster.

Terraform and Provider version

terraform 1.4.2

Affected resources and data sources

• Host
• Cluster

Terraform configuration

No response

Expected behavior

Applying the same plan shouldn't change the previously created host.

Actual behavior

Host was removed from the cluster

Steps to Reproduce

Run this twice, host is out of the cluster

terraform {
  required_providers {
    opennebula = {
      source = "OpenNebula/opennebula"
    }
  }
}

provider "opennebula" {
  endpoint = "http:/XXXXX:2633/RPC2"
  username = "XXXXX"
}

resource "opennebula_cluster" "this" {
  name = join("-", ["deployment_issue", "cluster"])
  tags = {
    deployment_id = "deployment_issue"
  }
}

resource "opennebula_host" "this" {
  name       = "deployment_issue-host"
  type       = "kvm"
  cluster_id = opennebula_cluster.this.id
  tags = {
    deployment_id = "deployment_issue"
  }
}

Debug output

No response

Panic output

No response

Important factoids

No response

References

No response

[vickmp]

Hi @tinova

After several attempts I can't reproduce the issue. When I try to add a Host to OpenNebula that can be monitored (assigning at the same time a Cluster as you show in your example) it always works fine no matter how many times I apply the changes.

However, if I apply exactly the file that you comment in the body of the issue, the behavior is as follows:

1. The gost is created and added to the cluster, but the host (deployment_issue-host) stays in error state as it cannot be monitored. This is to be expected since there is no host accessible via this hostname in my test lab:

   opennebula_host.this: Creating...
   opennebula_host.this: Still creating... [10s elapsed]
   ╷
   │ Error: Failed to wait host to be in MONITORED state
   │
   │   with opennebula_host.this,
   │   on site.tf line 31, in resource "opennebula_host" "this":
   │   31: resource "opennebula_host" "this" {
   │
   │ host (ID: 5): Host (ID:5) entered fail state, error: Thu Nov  2 11:29:29 2023 : Error monitoring Host development_host (5):
   ╵
   root@frontend:~/trrfrm# onehost list
     ID NAME                                                                CLUSTER    TVM      ALLOCATED_CPU      ALLOCATED_MEM STAT
      5 development_host                                                    deployment   0                  -                  - err

2. When running terraform apply again, the host is indeed destroyed, but it's recreated again where this new host is correctly added to the cluster.

   Terraform will perform the following actions:
   
     # opennebula_host.this is tainted, so must be replaced
   -/+ resource "opennebula_host" "this" {
         ~ default_tags = {} -> (known after apply)
         ~ id           = "6" -> (known after apply)
           name         = "development_host"
         ~ tags         = {
             ~ "deployment_id" = "" -> "deployment_issue"
           }
         ~ tags_all     = {
             - "deployment_id" = ""
           } -> (known after apply)
           # (2 unchanged attributes hidden)
       }
   
   Plan: 1 to add, 0 to change, 1 to destroy.
   
   opennebula_host.this: Destroying... [id=5]
   opennebula_host.this: Still destroying... [id=5, 10s elapsed]
   opennebula_host.this: Destruction complete after 10s
   opennebula_host.this: Creating...
   opennebula_host.this: Still creating... [10s elapsed]
   ╷
   │ Error: Failed to wait host to be in MONITORED state
   │
   │   with opennebula_host.this,
   │   on site.tf line 22, in resource "opennebula_host" "this":
   │   22: resource "opennebula_host" "this" {
   │
   │ host (ID: 6): Host (ID:6) entered fail state, error: Thu Nov  2 11:23:14 2023 : Error monitoring Host deployment_issue-host
   │ (6):
   ╵
   root@frontend:~/trrfrm# onehost list
     ID NAME                                                                CLUSTER    TVM      ALLOCATED_CPU      ALLOCATED_MEM STAT
      6 deployment_issue-host                                               deployment   0                  -                  - err

Isn't this the expected behavior? Maybe there is something I am not taking into account?

My test environement configuration:

• Terraform v1.6.3
• Terraform opennebula provider v1.3.1
• OpenNebula 6.8

[tinova]

closing until we get more info to reproduce it