fix(autoscale): honour --region instead of silently ignoring it

CHANGELOG.md

@@ -196,6 +196,25 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- **`spawn autoscale` honours `--region` instead of silently ignoring it**
+  (#774). The root `--region` flag — whose help says it "overrides
+  SPORE_REGION/AWS_REGION and the shared config" — overrode nothing: every
+  autoscale call site built its AWS config with a bare
+  `config.LoadDefaultConfig(ctx)`, taking the region from the ambient chain. It
+  was ignored in *both* directions, with `AWS_REGION` winning even when
+  `--region` contradicted it.
+  The symptoms: the groups listing failed with `ResourceNotFoundException` in any
+  region without a groups table; `autoscale health` reported a healthy group as
+  having no instances; and **`autoscale terminate` would find nothing to
+  terminate and report success while the instances kept billing**. It also made
+  #772's new coverage check report "no autoscale orchestrator runs in this
+  account" against an account that has one.
+  All **four** call sites now route through one resolver, which fixes the shared
+  *profile* they also ignored and uses the same precedence as every other command
+  (explicit region > shared config > ambient). Two of the four were found by the
+  new gate, not by me — `autoscale health` and `autoscale terminate` both had
+  `cfg, _ := config.LoadDefaultConfig(ctx)`, discarding the error as well.
+
 - **Three `pipeline-orchestrator` Makefile targets named a Lambda that does not
   exist** (#754). `make update-code`, `make logs` and `make invoke-test` all
   asked for `spawn-pipeline-orchestrator-production`, but `template.yaml` sets

cmd/autoscale_coverage.go

@@ -5,7 +5,6 @@
 	"fmt"
 	"io"
 
-	"github.com/aws/aws-sdk-go-v2/config"
 	spawnaws "github.com/spore-host/spawn/pkg/aws"
 )
 
@@ -26,7 +25,12 @@
 // turn a successful launch into a failure, so every error path here degrades to
 // either a hedged note or silence.
 func reportAutoscaleCoverage(ctx context.Context, w io.Writer) spawnaws.AutoscaleCoverage {
-	cfg, err := config.LoadDefaultConfig(ctx)
+	// The SAME resolver the rest of the command uses (spawn#774), so coverage
+	// cannot report on a different region than the one being operated on. It
+	// previously called config.LoadDefaultConfig directly and reported "no
+	// autoscale orchestrator runs in this account" against an account that has
+	// one, because the ambient region was us-west-2.
+	cfg, err := autoscaleConfig(ctx)
 	if err != nil {
 		// Nothing useful to say, and the caller's real work already succeeded.
 		return spawnaws.AutoscaleCoverage{Why: err.Error()}

cmd/autoscale_helpers.go

@@ -5,20 +5,46 @@
 	"fmt"
 
 	"github.com/aws/aws-sdk-go-v2/aws"
-	"github.com/aws/aws-sdk-go-v2/config"
 	"github.com/aws/aws-sdk-go-v2/service/cloudwatch"
 	"github.com/aws/aws-sdk-go-v2/service/dynamodb"
 	"github.com/aws/aws-sdk-go-v2/service/ec2"
 	"github.com/aws/aws-sdk-go-v2/service/lambda"
 	lambdaTypes "github.com/aws/aws-sdk-go-v2/service/lambda/types"
 	"github.com/aws/aws-sdk-go-v2/service/sqs"
 	"github.com/spore-host/spawn/pkg/autoscaler"
+	spawnaws "github.com/spore-host/spawn/pkg/aws"
 )
 
+// autoscaleConfig resolves the AWS config for every autoscale call site.
+//
+// This is spawn#774. Both sites used to call config.LoadDefaultConfig(ctx) with
+// no options, which took the region from the ambient chain and IGNORED the
+// root --region flag — a flag whose own help text says it "overrides
+// SPORE_REGION/AWS_REGION and the shared config". It overrode nothing:
+//
+//	$ spawn autoscale status --region us-east-1   # probed us-west-2
+//	$ AWS_REGION=us-east-1 … --region us-west-2   # used us-east-1
+//
+// Ignored in both directions, with the env winning over the flag that claims to
+// beat it. The visible symptoms were a DynamoDB scan failing with
+// ResourceNotFoundException in a region holding no groups table, and
+// triggerLambda invoking a function by name in a region where it may not exist.
+//
+// Routing through NewClientWithRegion fixes the shared PROFILE too, which these
+// sites also ignored, and uses the same resolution order as every other command
+// (explicit region > sporeconfig shared region > ambient).
+func autoscaleConfig(ctx context.Context) (aws.Config, error) {
+	client, err := spawnaws.NewClientWithRegion(ctx, spawnRegion)
+	if err != nil {
+		return aws.Config{}, fmt.Errorf("load AWS config: %w", err)
+	}
+	return client.Config(), nil
+}
+
 func getAutoscaler(ctx context.Context) (*autoscaler.AutoScaler, error) {
-	cfg, err := config.LoadDefaultConfig(ctx)
+	cfg, err := autoscaleConfig(ctx)
 	if err != nil {
-		return nil, fmt.Errorf("load AWS config: %w", err)
+		return nil, err
 	}
 
 	// Build table name with environment suffix
@@ -40,7 +66,7 @@
 }
 
 func triggerLambda(ctx context.Context, groupID string) error {
-	cfg, err := config.LoadDefaultConfig(ctx)
+	cfg, err := autoscaleConfig(ctx)
 	if err != nil {
 		return err
 	}

cmd/autoscale_lifecycle.go

@@ -6,7 +6,6 @@
 	"log"
 
 	"github.com/aws/aws-sdk-go-v2/aws"
-	"github.com/aws/aws-sdk-go-v2/config"
 	"github.com/aws/aws-sdk-go-v2/service/ec2"
 	"github.com/aws/aws-sdk-go-v2/service/ec2/types"
 	"github.com/spf13/cobra"
@@ -140,8 +139,16 @@
 		return fmt.Errorf("get group: %w", err)
 	}
 
-	// Terminate all instances
-	cfg, _ := config.LoadDefaultConfig(ctx)
+	// Terminate all instances.
+	//
+	// The region matters here more than anywhere else in this command (#774):
+	// with the wrong one this finds NOTHING to terminate and reports success
+	// while the instances keep billing. The error is handled rather than
+	// discarded for the same reason.
+	cfg, err := autoscaleConfig(ctx)
+	if err != nil {
+		return err
+	}
 	ec2Client := ec2.NewFromConfig(cfg)
 
 	result, err := ec2Client.DescribeInstances(ctx, &ec2.DescribeInstancesInput{

cmd/autoscale_region_test.go

@@ -0,0 +1,68 @@
+package cmd
+
+import (
+	"os"
+	"path/filepath"
+	"regexp"
+	"strings"
+	"testing"
+)
+
+// TestAutoscaleHonoursTheRegionFlag guards spawn#774.
+//
+// `spawn autoscale` advertised the root --region flag — whose help says it
+// "overrides SPORE_REGION/AWS_REGION and the shared config" — and ignored it.
+// Both call sites built their AWS config with a bare
+// config.LoadDefaultConfig(ctx), so the region came from the ambient chain and
+// the flag did nothing. The flag was ignored in BOTH directions: AWS_REGION won
+// even when --region contradicted it.
+//
+// Symptoms were a DynamoDB scan failing with ResourceNotFoundException against a
+// region with no groups table, triggerLambda invoking a function by name in a
+// region where it may not exist, and the #772 coverage check confidently
+// reporting "no autoscale orchestrator runs in this account" against an account
+// that has one.
+//
+// A source assertion rather than a behavioural one, deliberately: proving a
+// region reached the SDK needs either a live account or an injected client, and
+// neither exists on this path today. What is cheap and sufficient is forbidding
+// the construct that caused it — an unparameterised LoadDefaultConfig in the
+// autoscale call sites, which must go through autoscaleConfig instead.
+func TestAutoscaleHonoursTheRegionFlag(t *testing.T) {
+	// Matches a LoadDefaultConfig call with NO options — the broken form.
+	// `config.LoadDefaultConfig(ctx, config.WithRegion(r))` is fine and does not
+	// match, which is what distinguishes this from a blanket ban.
+	bare := regexp.MustCompile(`config\.LoadDefaultConfig\(\s*ctx\s*\)`)
+
+	files, err := filepath.Glob("autoscale*.go")
+	if err != nil {
+		t.Fatalf("glob: %v", err)
+	}
+	if len(files) == 0 {
+		t.Fatal("no autoscale*.go files found — this gate is matching nothing")
+	}
+
+	for _, f := range files {
+		if strings.HasSuffix(f, "_test.go") {
+			continue
+		}
+		b, err := os.ReadFile(f)
+		if err != nil {
+			t.Fatalf("read %s: %v", f, err)
+		}
+		// Strip comments so the explanation above (which names the broken call)
+		// does not satisfy the matcher. Three gates this session passed against
+		// their own prose; this one is written knowing that.
+		for _, line := range strings.Split(string(b), "\n") {
+			if strings.HasPrefix(strings.TrimSpace(line), "//") {
+				continue
+			}
+			if bare.MatchString(line) {
+				t.Errorf("%s: config.LoadDefaultConfig(ctx) with no options ignores the --region "+
+					"flag and the shared profile (spawn#774). Use autoscaleConfig(ctx), which "+
+					"routes through aws.NewClientWithRegion(ctx, spawnRegion).\n  %s",
+					f, strings.TrimSpace(line))
+			}
+		}
+	}
+}

cmd/autoscale_status.go

@@ -7,7 +7,6 @@
 	"time"
 
 	"github.com/aws/aws-sdk-go-v2/aws"
-	"github.com/aws/aws-sdk-go-v2/config"
 	"github.com/aws/aws-sdk-go-v2/service/dynamodb"
 	"github.com/aws/aws-sdk-go-v2/service/ec2"
 	"github.com/aws/aws-sdk-go-v2/service/ec2/types"
@@ -100,8 +99,13 @@
 		return fmt.Errorf("get group: %w", err)
 	}
 
-	// Get instances
-	cfg, _ := config.LoadDefaultConfig(ctx)
+	// Get instances. Region-resolved like every other autoscale call site
+	// (#774); a wrong region here reports a healthy group as having no
+	// instances.
+	cfg, err := autoscaleConfig(ctx)
+	if err != nil {
+		return err
+	}
 	ec2Client := ec2.NewFromConfig(cfg)
 
 	result, err := ec2Client.DescribeInstances(ctx, &ec2.DescribeInstancesInput{