Recent terraform fixes

playbooks/terraform.yml

@@ -1,6 +1,5 @@
 ---
 - name: Manage infrastructure lifecycle and SSH access with Terraform
   hosts: localhost
-  gather_facts: false
   roles:
     - role: terraform

scripts/check-ssh-key-migration.sh

@@ -0,0 +1,35 @@
+#!/bin/bash
+# SPDX-License-Identifier: copyleft-next-0.3.1
+#
+# Check if SSH keys need migration from old (unhashed) to new (hashed) paths.
+# This helps users upgrading from older kdevops versions that used a fixed
+# SSH key path to the new per-directory hashed paths.
+
+set -e
+
+TOPDIR_PATH="${1:-.}"
+HASH=$(echo "$TOPDIR_PATH" | sha256sum | cut -c1-8)
+
+OLD_KEY="$HOME/.ssh/kdevops_terraform"
+OLD_PUBKEY="$HOME/.ssh/kdevops_terraform.pub"
+NEW_KEY="$HOME/.ssh/kdevops_terraform_${HASH}"
+NEW_PUBKEY="$HOME/.ssh/kdevops_terraform_${HASH}.pub"
+
+# Only show notice if old key exists but new key doesn't
+if [ -f "$OLD_PUBKEY" ] && [ ! -f "$NEW_PUBKEY" ]; then
+	cat <<EOF
+--------------------------------------------------------------------------------
+NOTE: SSH key path has changed
+
+kdevops now uses directory-specific SSH key paths. An old-style key exists:
+  Old: $OLD_PUBKEY
+  New: $NEW_PUBKEY
+
+If you have RUNNING VMs that need the old key, migrate it:
+  mv "$OLD_KEY" "$NEW_KEY"
+  mv "$OLD_PUBKEY" "$NEW_PUBKEY"
+
+Otherwise, a new key will be generated automatically.
+--------------------------------------------------------------------------------
+EOF
+fi

scripts/terraform.Makefile

@@ -92,9 +92,14 @@ ifeq (y,$(CONFIG_TERRAFORM_SSH_CONFIG_GENKEY_OVERWRITE))
 DEFAULT_DEPS += remove-ssh-key
 endif
 
+DEFAULT_DEPS += ssh-key-migration-check
 DEFAULT_DEPS += $(KDEVOPS_SSH_PRIVKEY)
 endif # CONFIG_TERRAFORM_SSH_CONFIG_GENKEY
 
+PHONY += ssh-key-migration-check
+ssh-key-migration-check:
+	$(Q)$(TOPDIR)/scripts/check-ssh-key-migration.sh $(TOPDIR_PATH)
+
 ANSIBLE_EXTRA_ARGS += $(TERRAFORM_EXTRA_VARS)
 
 # Lambda Labs SSH key management

terraform/Kconfig.ssh

@@ -21,14 +21,17 @@ config TERRAFORM_SSH_CONFIG_USER
 
 config TERRAFORM_SSH_CONFIG_PUBKEY_FILE
 	string "File containing Ansible's ssh public key"
-	default "~/.ssh/kdevops_terraform_$(shell, echo $(TOPDIR_PATH) | sha256sum | cut -c1-8).pub" if TERRAFORM_LAMBDALABS
-	default "~/.ssh/kdevops_terraform.pub"
+	default "~/.ssh/kdevops_terraform_$(shell, echo $(TOPDIR_PATH) | sha256sum | cut -c1-8).pub"
 	help
 	  The filename of the file containing an ssh public key
 	  Ansible is to use to manage its target nodes. The
 	  matching private key should be located in a file using
 	  the same basename (without the ".pub").
 
+	  The filename includes an 8-character hash of the current
+	  directory path, allowing multiple kdevops installations to
+	  use separate SSH keys without conflicts.
+
 config TERRAFORM_SSH_CONFIG_GENKEY
 	bool "Should we create a new random key for you?"
 	default y

terraform/aws/scripts/aws_common.py

@@ -18,6 +18,12 @@
 from botocore.exceptions import ClientError, NoCredentialsError
 
 
+class AwsNotConfiguredError(Exception):
+    """Raised when AWS credentials are not available."""
+
+    pass
+
+
 def get_default_region():
     """
     Get the default AWS region from the ~/.aws/config file.
@@ -128,6 +134,30 @@ def handle_aws_credentials_error(quiet=False):
     return False
 
 
+def require_aws_credentials():
+    """
+    Require AWS credentials, raising an exception if not configured.
+
+    This function should be called early in main() to validate AWS
+    credentials. If AWS is not configured, it raises AwsNotConfiguredError
+    to let the caller decide how to handle it.
+
+    This centralizes the handling of missing AWS credentials and avoids
+    TOCTOU race conditions from manual file existence checks.
+
+    Returns:
+        dict: Caller identity information if credentials are valid
+
+    Raises:
+        AwsNotConfiguredError: If AWS credentials are not found
+    """
+    try:
+        sts = boto3.client("sts")
+        return sts.get_caller_identity()
+    except NoCredentialsError as e:
+        raise AwsNotConfiguredError("AWS credentials not found") from e
+
+
 def get_all_regions(quiet=False):
     """
     Retrieve the list of all AWS regions.

terraform/aws/scripts/gen_kconfig_ami

@@ -14,15 +14,15 @@ from collections import defaultdict
 from datetime import datetime, timedelta
 from concurrent.futures import ThreadPoolExecutor, as_completed
 
-import boto3
-from botocore.exceptions import ClientError, NoCredentialsError
+from botocore.exceptions import ClientError
 
 from aws_common import (
+    AwsNotConfiguredError,
     get_default_region,
     get_jinja2_environment,
     create_ec2_client,
     handle_aws_client_error,
-    handle_aws_credentials_error,
+    require_aws_credentials,
 )
 
 
@@ -257,9 +257,6 @@ def discover_ami_patterns(
 
         return discovered_patterns
 
-    except NoCredentialsError:
-        handle_aws_credentials_error(quiet)
-        return {}
     except ClientError as e:
         handle_aws_client_error(e, f"discovering AMI patterns for {owner_name}", quiet)
         return {}
@@ -822,6 +819,14 @@ def main():
             output_owners_raw(owners, args.quiet)
         return
 
+    # Allow make dynconfig to succeed without AWS credentials
+    try:
+        require_aws_credentials()
+    except AwsNotConfiguredError:
+        if not args.quiet:
+            print("AWS not configured - skipping (optional)", file=sys.stderr)
+        sys.exit(0)
+
     if args.region:
         region = args.region
     else:

terraform/aws/scripts/gen_kconfig_instance

@@ -9,14 +9,12 @@ construct the "instance" Kconfig menu.
 import sys
 import argparse
 
-from botocore.exceptions import ClientError, NoCredentialsError
-
 from aws_common import (
+    AwsNotConfiguredError,
     get_default_region,
     get_all_instance_types,
     get_jinja2_environment,
-    handle_aws_client_error,
-    handle_aws_credentials_error,
+    require_aws_credentials,
 )
 
 
@@ -316,6 +314,14 @@ def main():
     """Main function to run the program."""
     args = parse_arguments()
 
+    # Allow make dynconfig to succeed without AWS credentials
+    try:
+        require_aws_credentials()
+    except AwsNotConfiguredError:
+        if not args.quiet:
+            print("AWS not configured - skipping (optional)", file=sys.stderr)
+        sys.exit(0)
+
     if args.region:
         region = args.region
     else:

terraform/aws/scripts/gen_kconfig_location

@@ -10,17 +10,15 @@ import sys
 import argparse
 
 from concurrent.futures import ThreadPoolExecutor, as_completed
-from botocore.exceptions import ClientError, NoCredentialsError
 
 from aws_common import (
+    AwsNotConfiguredError,
     get_default_region,
     get_all_regions,
     get_region_availability_zones,
     get_jinja2_environment,
     get_region_kconfig_name,
-    create_ec2_client,
-    handle_aws_client_error,
-    handle_aws_credentials_error,
+    require_aws_credentials,
 )
 
 
@@ -214,6 +212,14 @@ def main():
     """Main function to run the program."""
     args = parse_arguments()
 
+    # Allow make dynconfig to succeed without AWS credentials
+    try:
+        require_aws_credentials()
+    except AwsNotConfiguredError:
+        if not args.quiet:
+            print("AWS not configured - skipping (optional)", file=sys.stderr)
+        sys.exit(0)
+
     if not args.quiet:
         print("Fetching list of all AWS regions...", file=sys.stderr)
     regions = get_all_regions()

terraform/azure/scripts/azure_common.py

@@ -17,6 +17,12 @@
 from jinja2 import Environment, FileSystemLoader
 
 
+class AzureNotConfiguredError(Exception):
+    """Raised when Azure credentials are not available."""
+
+    pass
+
+
 def get_default_region():
     """
     Get the default Azure region from Azure configuration.
@@ -371,3 +377,45 @@ def exit_on_empty_result(result, context, quiet=False):
             )
             print("Run 'az login' to authenticate with Azure.", file=sys.stderr)
         sys.exit(1)
+
+
+def require_azure_credentials():
+    """
+    Require Azure credentials, raising an exception if not configured.
+
+    This function should be called early in main() to validate Azure
+    credentials. If Azure is not configured, it raises AzureNotConfiguredError
+    to let the caller decide how to handle it.
+
+    This centralizes the handling of missing Azure credentials and avoids
+    TOCTOU race conditions from manual file existence checks.
+
+    Returns:
+        str: Subscription ID if credentials are valid
+
+    Raises:
+        AzureNotConfiguredError: If Azure credentials are not found
+    """
+    try:
+        from azure.common.credentials import get_cli_profile
+
+        profile = get_cli_profile()
+        credentials, subscription_id, _ = profile.get_login_credentials(
+            resource="https://management.azure.com"
+        )
+        return subscription_id
+    except ImportError as e:
+        raise AzureNotConfiguredError("Azure SDK not installed") from e
+    except Exception as e:
+        # Only treat as "not configured" if it looks like an auth/login issue
+        error_msg = str(e).lower()
+        auth_indicators = [
+            "login",
+            "logged in",
+            "authenticate",
+            "credential",
+            "az login",
+        ]
+        if any(phrase in error_msg for phrase in auth_indicators):
+            raise AzureNotConfiguredError("Azure credentials not found") from e
+        raise

terraform/azure/scripts/gen_kconfig_image

@@ -55,11 +55,12 @@ from collections import defaultdict
 from functools import lru_cache
 
 from azure_common import (
+	AzureNotConfiguredError,
 	get_default_region,
 	get_jinja2_environment,
-	get_all_regions,
 	get_region_kconfig_name,
 	get_all_offers_and_skus,
+	require_azure_credentials,
 )
 
 
@@ -730,6 +731,14 @@ def main():
 		output_publishers_raw(args.quiet)
 		return
 
+	# Allow make dynconfig to succeed without Azure credentials
+	try:
+		require_azure_credentials()
+	except AzureNotConfiguredError:
+		if not args.quiet:
+			print("Azure not configured - skipping (optional)", file=sys.stderr)
+		sys.exit(0)
+
 	publishers = get_known_publishers()
 
 	# Filter to specific publisher if requested

terraform/azure/scripts/gen_kconfig_location

@@ -26,11 +26,12 @@ import sys
 import argparse
 
 from azure_common import (
+    AzureNotConfiguredError,
     get_default_region,
     get_all_regions,
     get_jinja2_environment,
     get_region_kconfig_name,
-    exit_on_empty_result,
+    require_azure_credentials,
 )
 
 
@@ -191,11 +192,18 @@ def main():
     """Main function to run the program."""
     args = parse_arguments()
 
+    # Allow make dynconfig to succeed without Azure credentials
+    try:
+        require_azure_credentials()
+    except AzureNotConfiguredError:
+        if not args.quiet:
+            print("Azure not configured - skipping (optional)", file=sys.stderr)
+        sys.exit(0)
+
     if not args.quiet:
         print("Fetching list of all Azure regions...", file=sys.stderr)
 
     regions = get_all_regions(args.quiet)
-    exit_on_empty_result(regions, "Azure region query", args.quiet)
 
     if args.regions:
         if args.format == "kconfig":

terraform/azure/scripts/gen_kconfig_size

@@ -39,11 +39,12 @@ import os
 import yaml
 
 from azure_common import (
+    AzureNotConfiguredError,
     get_default_region,
     get_all_regions,
     get_jinja2_environment,
     get_vm_sizes_and_skus,
-    exit_on_empty_result,
+    require_azure_credentials,
 )
 
 
@@ -602,21 +603,27 @@ def main():
     """Main function to run the program."""
     args = parse_arguments()
 
+    # Allow make dynconfig to succeed without Azure credentials
+    try:
+        require_azure_credentials()
+    except AzureNotConfiguredError:
+        if not args.quiet:
+            print("Azure not configured - skipping (optional)", file=sys.stderr)
+        sys.exit(0)
+
     # Determine which regions to query
     if args.region:
         # Query specific region only
         regions = [args.region]
     elif args.all_regions:
         # Query all regions
         regions = get_all_regions(args.quiet)
-        exit_on_empty_result(regions, "Azure region query", args.quiet)
     else:
         # Query default region only
         regions = [get_default_region()]
 
     # Get VM sizes and capabilities in a single API call
     sizes, sku_capabilities = get_all_vm_sizes_and_capabilities(regions, args.quiet)
-    exit_on_empty_result(sizes, "Azure VM size query", args.quiet)
 
     if args.families:
         output_families_raw(sizes, args.quiet)