diff --git a/cloud/tq/queuehandler.go b/cloud/tq/queuehandler.go
index ae5a5e8c..7072cd12 100644
--- a/cloud/tq/queuehandler.go
+++ b/cloud/tq/queuehandler.go
@@ -201,6 +201,9 @@ func (qh *ChannelQueueHandler) handleLoop(next api.TaskPipe, bucketOpts ...optio
 			}
 		} else {
 			log.Println("No task files")
+			task.Queue = ""
+			task.Update(state.Done)
+			task.Delete()
 		}
 	}
 	log.Println(qh.Queue, "waiting for deduper to close")
diff --git a/dispatch/deduphandler.go b/dispatch/deduphandler.go
index 6147952c..cb1d18e2 100644
--- a/dispatch/deduphandler.go
+++ b/dispatch/deduphandler.go
@@ -222,12 +222,13 @@ func (dh *DedupHandler) handleLoop(opts ...option.ClientOption) {
 // feeding channel is closed, and processing is complete.
 func NewDedupHandler(opts ...option.ClientOption) *DedupHandler {
 	project := os.Getenv("PROJECT")
+	dataset := os.Getenv("DATASET")
 	// When running in prod, the task files and queues are in mlab-oti, but the destination
 	// BigQuery tables are in measurement-lab.
-	if project == "mlab-oti" {
+	// However, for sidestream private tables, we leave them in mlab-oti
+	if project == "mlab-oti" && dataset != "private" {
 		project = "measurement-lab" // destination for production tables.
 	}
-	dataset := os.Getenv("DATASET")
 	msg := make(chan state.Task)
 	rsp := make(chan error)
 	dh := DedupHandler{project, dataset, msg, rsp}
@@ -312,4 +313,3 @@ func Dedup(dsExt *bqext.Dataset, src string, destTable *bigquery.Table) (*bigque
 	}
 	return job, nil
 }
-
diff --git a/k8s/data-processing-cluster/deployments/etl-gardener.yml b/k8s/data-processing-cluster/deployments/etl-gardener.yml
index ecd605d7..49dd19a6 100644
--- a/k8s/data-processing-cluster/deployments/etl-gardener.yml
+++ b/k8s/data-processing-cluster/deployments/etl-gardener.yml
@@ -44,13 +44,13 @@ spec:
         - name: PROJECT
           value: {{GCLOUD_PROJECT}}
         - name: TASKFILE_BUCKET
-          value: "archive-{{GCLOUD_PROJECT}}"
+          value: "scraper-{{GCLOUD_PROJECT}}"  # NOTE: if we start deleting unembargoed files from scraper, this will no longer work.
         - name: START_DATE
-          value: "20180301"
+          value: "20170601"
         - name: EXPERIMENTS
-          value: "sidestream,ndt"  # For example "ndt,sidestream,switch"
+          value: "sidestream"  # For example "ndt,sidestream,switch"
         - name: DATASET
-          value: batch
+          value: private
         - name: FINAL_DATASET
           value: ""  # e.g. base_tables
         - name: QUEUE_BASE