snapshot-agent: move the GPU-CR sweeper to the utils package

backends/gc.go -> utils/gpu-cr_utils.go (tests likewise), StartGC ->
StartGPUCRSweeper. The sweeper is GPU-CR hygiene shared by any GPU-CR-driven
backend, not backend logic itself, so it lives with the other node-level
helpers.

Author: Edwinhr716

cmd/snapshot-agent/main.go

@@ -26,6 +26,7 @@ import (
 	"github.com/llm-d-incubation/llm-d-rl-time-slicing/pkg/snapshot-agent/backends"
 	"github.com/llm-d-incubation/llm-d-rl-time-slicing/pkg/snapshot-agent/features"
 	"github.com/llm-d-incubation/llm-d-rl-time-slicing/pkg/snapshot-agent/server"
+	"github.com/llm-d-incubation/llm-d-rl-time-slicing/pkg/snapshot-agent/utils"
 )
 
 func main() {
@@ -87,8 +88,7 @@ func main() {
 	}
 
 	// GPU-CR housekeeping runs only when the shared checkpoint dir is
-	// configured (the Helm chart sets EXPORT_FILE_PATH iff a GPU-CR-driven
-	// backend is enabled), keeping CUDA/app-only deployments untouched.
+	// configured, keeping CUDA/app-only deployments untouched.
 	if ctlDir := os.Getenv("EXPORT_FILE_PATH"); ctlDir != "" {
 		// The dir must be writable by the (unprivileged) GPU-CR workloads
 		// that mmap their dump buffers in it.
@@ -101,7 +101,7 @@ func main() {
 		}
 		// Sweep stale GPU-CR artifacts: a leaked dump pins its full extent
 		// in shm/hugetlbfs even after the owning process dies.
-		backends.StartGC(ctx, ctlDir, 10*time.Minute)
+		utils.StartGPUCRSweeper(ctx, ctlDir, 10*time.Minute)
 	}
 
 	slog.InfoContext(ctx, "Starting Snapshot Agent",

pkg/snapshot-agent/backends/gc.go pkg/snapshot-agent/utils/gpu-cr_utils.gopkg/snapshot-agent/backends/gc.go renamed to pkg/snapshot-agent/utils/gpu-cr_utils.go

@@ -1,4 +1,4 @@
-package backends
+package utils
 
 import (
 	"context"
@@ -29,8 +29,9 @@ var (
 // mmap'd them yet (files appear before the mapping does).
 const gcMinAge = 5 * time.Minute
 
-// StartGC sweeps stale GPU-CR artifacts at startup and every interval.
-func StartGC(ctx context.Context, ctlDir string, interval time.Duration) {
+// StartGPUCRSweeper sweeps stale GPU-CR artifacts at startup and every
+// interval.
+func StartGPUCRSweeper(ctx context.Context, ctlDir string, interval time.Duration) {
 	go func() {
 		sweep(ctlDir)
 		t := time.NewTicker(interval)

…pshot-agent/backends/gc_internal_test.go …napshot-agent/utils/gpu-cr_utils_test.gopkg/snapshot-agent/backends/gc_internal_test.go renamed to pkg/snapshot-agent/utils/gpu-cr_utils_test.go pkg/snapshot-agent/backends/gc_internal_test.go renamed to pkg/snapshot-agent/utils/gpu-cr_utils_test.go

@@ -1,4 +1,4 @@
-package backends
+package utils
 
 import (
 	"os"