Bug: Panic when removing resource_monitoring block from coder_agent resource

[blink-so]

Description

Removing the resource_monitoring block from a coder_agent resource causes a panic with "index out of range [0] with length 0" error when starting workspaces. This appears to be a long-standing issue that affects workspace functionality when users attempt to disable resource monitoring.

Note: The issue occurs both when completely removing the block AND when replacing the content with empty braces {}.

Expected Behavior

• Removing the resource_monitoring block should either:
  1. Allow the workspace to start normally without monitoring (if this is intended to be optional)
  2. Provide a clear validation error explaining that resource_monitoring is required
• Empty resource_monitoring {} blocks should either be handled gracefully or validated properly

Actual Behavior

• Workspace fails to start with a panic
• Error message: The plugin.(*GRPCProvider).PlanResourceChange request was cancelled.
• Stack trace shows: panic: runtime error: index out of range [0] with length 0

Steps to Reproduce

1. Create a template with a coder_agent resource that includes a resource_monitoring block
2. Successfully deploy workspaces using this template
3. Either:
   • Remove the entire resource_monitoring block from the coder_agent resource, OR
   • Replace the resource_monitoring content with empty braces: resource_monitoring {}
4. Update the template
5. Attempt to start a workspace

Environment

• Provider Version: v2.10.0 (also reproduced in v2.8.0 and v2.4.0)
• Test platform: Coder 2.25.1

Example Configuration

Working configuration:

resource "coder_agent" "main" {
  os   = "linux"
  arch = "amd64"

  resources_monitoring {
    memory {
      enabled   = true
      threshold = 90
    }
    volume {
      path      = "/home/coder"
      enabled   = true
      threshold = 80
    }
  }
}

Failing configurations:

Option 1 - Completely removed:

resource "coder_agent" "main" {
  os   = "linux"
  arch = "amd64"
  # resource_monitoring block removed entirely
}

Option 2 - Empty block:

resource "coder_agent" "main" {
  os   = "linux"
  arch = "amd64"
  
  resources_monitoring {
    # empty block - also causes panic
  }
}

Additional Context

• This issue has been consistent across multiple provider versions (v2.4.0, v2.8.0, v2.10.0)
• The panic occurs at runtime when starting workspaces, not during template validation
• Template can be uploaded successfully, but workspace creation fails
• Rolling back the provider version does not resolve the issue once the resource_monitoring block is removed or emptied
• The error specifically affects line 86 in the customer's template (the coder_agent resource)
• Both complete removal and empty block replacement ({}) are equally destructive

Customer Impact

• Prevents users from disabling resource monitoring when desired
• Causes workspace startup failures
• Requires keeping resource monitoring enabled with actual content even when not needed
• No graceful way to disable monitoring functionality

Suggested Solutions

1. If resource_monitoring is optional: Fix the panic by properly handling the case where the block is absent or empty
2. If resource_monitoring is required: Add proper validation to prevent template upload/update when the block is missing or empty
3. Feature enhancement: Allow explicit disabling of resource monitoring with a boolean flag or handle empty blocks gracefully
4. Validation improvement: Provide clear error messages during template validation rather than runtime panics

Classification

This issue sits between a bug report and feature request:

• Bug aspect: The panic and poor error handling when the block is removed or emptied
• Feature aspect: The ability to optionally disable resource monitoring

[bjornrobertsson]

### Error log

Error: Plugin did not respond
on main.tf line 86, in resource "coder_agent" "main":
  86: resource "coder_agent" "main" {
The plugin encountered an error, and failed to respond to the plugin.(*GRPCProvider).PlanResourceChange call. The plugin logs may contain more details.
Warning: Failed to decode resource from state
Error decoding "coder_app.code-server" from prior state: unsupported attribute "group"
Stack trace from the terraform-provider-coder_v2.2.0 plugin:
panic: runtime error: index out of range [0] with length 0
goroutine 67 [running]:
github.com/coder/terraform-provider-coder/v2/provider.agentResource.func5({0xe3b000?, 0xc0005725d0?}, 0xc0003001c0, {0xd?, 0xd0bc84?})
	github.com/coder/terraform-provider-coder/v2/provider/agent.go:373 +0x73e
github.com/hashicorp/terraform-plugin-sdk/v2/helper/schema.schemaMap.Diff(0xc00038b710, {0xe3b000, 0xc0005725d0}, 0xc0000a5930, 0xc0000a1d60, 0xd5af20, {0xc275a0, 0xc000454e10}, 0x0)
	github.com/hashicorp/terraform-plugin-sdk/v2@v2.36.0/helper/schema/schema.go:710 +0x4b4
github.com/hashicorp/terraform-plugin-sdk/v2/helper/schema.(*Resource).SimpleDiff(0xe3b230?, {0xe3b000?, 0xc0005725d0?}, 0xc0000a5930, 0xc000572600?, {0xc275a0?, 0xc000454e10?})
	github.com/hashicorp/terraform-plugin-sdk/v2@v2.36.0/helper/schema/resource.go:1022 +0xdb
github.com/hashicorp/terraform-plugin-sdk/v2/helper/schema.(*GRPCProviderServer).PlanResourceChange(0xc00038c750, {0xe3b000?, 0xc000572510?}, 0xc0000a0e60)
	github.com/hashicorp/terraform-plugin-sdk/v2@v2.36.0/helper/schema/grpc_provider.go:890 +0xbce
github.com/hashicorp/terraform-plugin-go/tfprotov5/tf5server.(*server).PlanResourceChange(0xc0002a3180, {0xe3b000?, 0xc00045bef0?}, 0xc0001b3d80)
	github.com/hashicorp/terraform-plugin-go@v0.26.0/tfprotov5/tf5server/server.go:826 +0x3f0
github.com/hashicorp/terraform-plugin-go/tfprotov5/internal/tfplugin5._Provider_PlanResourceChange_Handler({0xce39e0, 0xc0002a3180}, {0xe3b000, 0xc00045bef0}, 0xc0001b3d00, 0x0)
	github.com/hashicorp/terraform-plugin-go@v0.26.0/tfprotov5/internal/tfplugin5/tfplugin5_grpc.pb.go:593 +0x1a6
google.golang.org/grpc.(*Server).processUnaryRPC(0xc000256200, {0xe3b000, 0xc00045be90}, 0xc000135560, 0xc00038bbf0, 0x13b28a0, 0x0)
	google.golang.org/grpc@v1.69.4/server.go:1392 +0xfd2
google.golang.org/grpc.(*Server).handleStream(0xc000256200, {0xe3b658, 0xc0004fa000}, 0xc000135560)
	google.golang.org/grpc@v1.69.4/server.go:1802 +0xbaa
google.golang.org/grpc.(*Server).serveStreams.func2.1()
	google.golang.org/grpc@v1.69.4/server.go:1030 +0x8b
created by google.golang.org/grpc.(*Server).serveStreams.func2 in goroutine 36
	google.golang.org/grpc@v1.69.4/server.go:1041 +0x125
Error: The terraform-provider-coder_v2.2.0 plugin crashed!
This is always indicative of a bug within the plugin. It would be immensely
helpful if you could report the crash with the plugin's maintainers so that it
can be fixed. The output above should help diagnose the issue.