diff --git a/api/v1/pipelinerun_types.go b/api/v1/pipelinerun_types.go
index a29abdc9..f6683f1e 100644
--- a/api/v1/pipelinerun_types.go
+++ b/api/v1/pipelinerun_types.go
@@ -17,9 +17,11 @@ limitations under the License.
 package v1
 
 import (
+	"fmt"
 	opsconstants "github.com/shaowenchen/ops/pkg/constants"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/types"
+	"time"
 )
 
 // EDIT THIS FILE!  THIS IS SCAFFOLDING FOR YOU TO OWN!
@@ -120,6 +122,14 @@ func (obj *PipelineRun) CopyWithOutVersion() *PipelineRun {
 	}
 }
 
+func (obj *PipelineRun) SetEnv() *PipelineRun {
+	if obj.Spec.Variables == nil {
+		obj.Spec.Variables = make(map[string]string)
+	}
+	obj.Spec.Variables["TIME"] = fmt.Sprintf("%d", time.Now().UnixMicro())
+	return obj
+}
+
 func NewPipelineRun(p *Pipeline) *PipelineRun {
 	if p == nil {
 		return &PipelineRun{}
diff --git a/controllers/pipelinerun_controller.go b/controllers/pipelinerun_controller.go
index 9ddde69a..4b7823ec 100644
--- a/controllers/pipelinerun_controller.go
+++ b/controllers/pipelinerun_controller.go
@@ -99,6 +99,8 @@ func (r *PipelineRunReconciler) Reconcile(ctx context.Context, req ctrl.Request)
 	if opsconstants.IsFinishedStatus(pr.Status.RunStatus) {
 		return ctrl.Result{}, nil
 	}
+	// insert env
+	pr.SetEnv()
 	// if is others cluster, send and just sync status
 	cluster := r.isOtherCluster(pr)
 	if cluster != nil {
diff --git a/controllers/taskrun_controller.go b/controllers/taskrun_controller.go
index 2d5b3264..d2c273e6 100644
--- a/controllers/taskrun_controller.go
+++ b/controllers/taskrun_controller.go
@@ -24,7 +24,6 @@ import (
 	"time"
 
 	"github.com/google/go-cmp/cmp"
-	"github.com/google/uuid"
 	cron "github.com/robfig/cron/v3"
 	opsv1 "github.com/shaowenchen/ops/api/v1"
 	opsconstants "github.com/shaowenchen/ops/pkg/constants"
@@ -219,6 +218,7 @@ func (r *TaskRunReconciler) run(logger *opslog.Logger, ctx context.Context, t *o
 		}
 	} else {
 		cluster := opsv1.NewCurrentCluster()
+		logger.Info.Println(fmt.Sprintf("run task %s on cluster %s", t.GetUniqueKey(), cluster.Name))
 		err = r.runTaskOnKube(cliLogger, ctx, t, tr, &cluster)
 		if err != nil {
 			logger.Error.Println(err)
@@ -252,7 +252,6 @@ func (r *TaskRunReconciler) runTaskOnHost(logger *opslog.Logger, ctx context.Con
 	vars["NAMESPACE"] = tr.Namespace
 	vars["OPSSERVER_ENDPOINT"] = r.getOpsServerEndpoint(t.Namespace)
 	vars["EVENT_CLUSTER"] = opsconstants.GetEnvEventCluster()
-	vars["UUID"] = uuid.New().String()
 
 	// insert host labels
 	for k, v := range h.ObjectMeta.Labels {
@@ -287,11 +286,11 @@ func (r *TaskRunReconciler) runTaskOnKube(logger *opslog.Logger, ctx context.Con
 		return err
 	}
 	// if find host in cluster, and can connect
-	host, _ := kc.GetHost(opsconstants.OpsNamespace, tr.GetHost(t))
-	if host != nil {
-		logger.Info.Println("use host credentials to run cluster task " + tr.Name)
-		return r.runTaskOnHost(logger, ctx, *kc.OpsClient, t, tr, host)
-	}
+	// host, _ := kc.GetHost(opsconstants.OpsNamespace, tr.GetHost(t))
+	// if host != nil {
+	// 	logger.Info.Println("use host credentials to run cluster task " + tr.Name)
+	// 	return r.runTaskOnHost(logger, ctx, *kc.OpsClient, t, tr, host)
+	// }
 	// else use pod to run task
 	// build options
 	hostStr := tr.GetHost(t)
@@ -323,7 +322,6 @@ func (r *TaskRunReconciler) runTaskOnKube(logger *opslog.Logger, ctx context.Con
 		vars["OPSSERVER_ENDPOINT"] = r.getOpsServerEndpoint(t.Namespace)
 		vars["TASK"] = t.Name
 		vars["TASKRUN"] = tr.Name
-		vars["UUID"] = uuid.New().String()
 		opstask.RunTaskOnKube(logger, t, tr, kc, &node,
 			opsoption.TaskOption{
 				Variables: vars,
diff --git a/tasks/collect-gpu-log.yaml b/tasks/collect-gpu-log.yaml
index e6ac7cd9..d55b7153 100644
--- a/tasks/collect-gpu-log.yaml
+++ b/tasks/collect-gpu-log.yaml
@@ -25,17 +25,16 @@ spec:
   steps:
     - name: run col_gpu_log.sh
       content: |
-        mkdir -p /tmp/collect-gpu-log || true
-        rm -rf /tmp/collect-gpu-log/*
-        cd /tmp/collect-gpu-log
+        mkdir -p /tmp/collect-gpu-log-${TIME}
+        cd /tmp/collect-gpu-log-${TIME}
         curl -sfL https://ghp.ci/https://raw.githubusercontent.com/shaowenchen/hubimage/main/ai/col_gpu_log.sh | bash
-        mv *.tar.gz collect-gpu-log-${HOSTNAME}-${UUID}.tar.gz
+        mv *.tar.gz collect-gpu-log-${HOSTNAME}-${TIME}.tar.gz
     - name: upload to s3
-      remotefile: s3://collect-gpu-log/collect-gpu-log-${HOSTNAME}-${UUID}.tar.gz
-      localfile: /tmp/collect-gpu-log/collect-gpu-log-${HOSTNAME}-${UUID}.tar.gz
+      remotefile: s3://collect-gpu-log/collect-gpu-log-${HOSTNAME}-${TIME}.tar.gz
+      localfile: /tmp/collect-gpu-log/collect-gpu-log-${HOSTNAME}-${TIME}.tar.gz
       direction: upload
     - name: get 2h shared link
       content: |
         curl -sfL https://ks3util-version-update.ks3-cn-beijing.ksyuncs.com/2.6.0/ks3util-linux-amd64 -o /usr/local/bin/ks3util >/dev/null
         chmod +x /usr/local/bin/ks3util
-        /usr/local/bin/ks3util sign ks3://${bucket}/collect-gpu-log/collect-gpu-log-${HOSTNAME}-${UUID}.tar.gz --endpoint ${endpoint} --timeout 7200 -i ${ak} -k ${sk}
+        /usr/local/bin/ks3util sign ks3://${bucket}/collect-gpu-log/collect-gpu-log-${HOSTNAME}-${TIME}.tar.gz --endpoint ${endpoint} --timeout 7200 -i ${ak} -k ${sk}