Bump gomaasclient & gomaasapi to avoid "net/http: cannot rewind body after connection loss" on terraform plan when TLS is enabled

[tnorris]

Howdy,

I am trying to use the MAAS terraform provider to spin up 8 machines and 8 instances. After enabling TLS on MAAS, the provider will sometimes throw a net/http error during plan; it looks like this:

module.maas-glue.maas_machine.fasrc_cluster_group["holyoke2c04311"]: Refreshing state... [id=bn7q4w]
module.maas-glue.maas_machine.fasrc_cluster_group["holyoke2c04312"]: Refreshing state... [id=cdhtnc]
module.maas-glue.maas_machine.fasrc_cluster_group["holyoke2a24216"]: Refreshing state... [id=q3rhqq]
module.maas-glue.maas_machine.fasrc_cluster_group["holyoke2c04316"]: Refreshing state... [id=pfqtsp]
module.maas-glue.maas_machine.fasrc_cluster_group["holyoke2c04314"]: Refreshing state... [id=dmtf4s]
module.maas-glue.maas_machine.fasrc_cluster_group["holyoke2c04310"]: Refreshing state... [id=pba8tg]
module.maas-glue.maas_machine.fasrc_cluster_group["holyokea24214"]: Refreshing state... [id=ww7fpr]
╷
│ Error: Get "https://holy-maas-region01.rc.fas.harvard.edu:5443/MAAS/api/2.0/machines/cdhtnc/": net/http: cannot rewind body after connection loss
│
│   with module.maas-glue.maas_machine.fasrc_cluster_group["holyoke2c04312"],
│   on ../modules/maas-glue/main.tf line 79, in resource "maas_machine" "fasrc_cluster_group":
│   79: resource "maas_machine" "fasrc_cluster_group" {
│

Ideally whatever is between Terraform and MaaS isn't hiccuping, but I was able to hack around this with a good ole retry loop (n.b.: I have no idea what I'm doing with golang):

diff --git i/maas/resource_maas_instance.go w/maas/resource_maas_instance.go
index d01b128..5d10a0b 100644
--- i/maas/resource_maas_instance.go
+++ w/maas/resource_maas_instance.go
@@ -3,6 +3,7 @@ package maas
 import (
    "context"
    "fmt"
+   "github.com/hashicorp/terraform-plugin-log/tflog"

    "github.com/hashicorp/terraform-plugin-sdk/v2/diag"
    "github.com/hashicorp/terraform-plugin-sdk/v2/helper/schema"
@@ -219,9 +220,23 @@ func resourceInstanceRead(ctx context.Context, d *schema.ResourceData, m interfa
    client := m.(*client.Client)

    // Get MAAS machine
-   machine, err := client.Machine.Get(d.Id())
-   if err != nil {
-       return diag.FromErr(err)
+   machine := *new(*entity.Machine)
+   err := *new(error)
+
+   failures := 0
+
+   for {
+       machine, err = client.Machine.Get(d.Id())
+       if err != nil {
+           failures++
+           tflog.Error(ctx, "Retrying a machine get")
+           if failures > 5 {
+               return diag.FromErr(err)
+
+           }
+           continue
+       }
+       break
    }
    // Set Terraform state
    ipAddresses := make([]string, len(machine.IPAddresses))

The only other encounter with this issue I could find was on a deleted bug in launchpad

I suspect there's A Better Way to retry, but I went for the quick and dirty.

[tnorris]

Or maybe this isn't retries, it's an empty response from MaaS? https://github.com/juju/gomaasapi/pull/98/files

[tnorris]

This can be fixed by bumping gomaasapi in gomaasclient, and then by bumping gomaasapi in terraform-provider-maas

    github.com/juju/gomaasapi/v2 v2.0.1 // indirect