diff --git a/apps/taskmasters/flowlord/README.md b/apps/taskmasters/flowlord/README.md
index 0e1c342f..5edc981d 100644
--- a/apps/taskmasters/flowlord/README.md
+++ b/apps/taskmasters/flowlord/README.md
@@ -17,23 +17,24 @@ template = "?date={yyyy}-{mm}-{dd}"
 [[Phase]]
 task = "task1"
 dependsOn = "topic_name"
-rule = ""
+rule = "retry_delay=30s"
 retry = 3
 template = "?hour={yyyy}-{mm}-{dd}T{hh}"
 ```
 
 ### Phase 
 
- - task: the name of the topic this task will be sent to. It is also the unique name of the task. In Addition a job of a task can be added in the task name using a colon (:) as a separator (task:job)
- - dependsOn: 
+ - **task**: the name of the topic this task will be sent to. It is also the unique name of the task. In Addition a job of a task can be added in the task name using a colon (:) as a separator (task:job)
+ - **dependsOn**: 
    - the name of the parent task
    - this task will start after the parent task has completed successfully
    - if left blank this tasks will only start based on the rule
- - rule: rules on about the tasks that are encoded as query params 
-   - cron: schedule the task based on the cron pattern (see scheduling)
-   - require: used in a child task saying to only start task if value is present
- - retry: the number of times a task is retried before being sent to failed_tasks
- - template: a URL string that is parsed and put into the task's info string when created
+ - **rule**: rules on about the tasks that are encoded as query params 
+   - _cron_: schedule the task based on the cron pattern (see scheduling)
+   - _require_: used in a child task saying to only start task if value is present
+   - _retry_delay_: duration to wait before retrying the task
+ - **retry**: the number of times a task is retried before being sent to failed_tasks
+ - **template**: a URL string that is parsed and put into the task's info string when created
 
 ### Template 
 templating is used to create dynamic tasks based on the time run or previous jobs run. templates are designated with surrounding brackets `{}`
diff --git a/apps/taskmasters/flowlord/taskmaster.go b/apps/taskmasters/flowlord/taskmaster.go
index 14695bfc..bf69f3ba 100644
--- a/apps/taskmasters/flowlord/taskmaster.go
+++ b/apps/taskmasters/flowlord/taskmaster.go
@@ -5,6 +5,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"log"
+	"math/rand"
 	"net/url"
 	"regexp"
 	"strconv"
@@ -24,6 +25,10 @@ import (
 	"github.com/pcelvng/task-tools/workflow"
 )
 
+func init() {
+	rand.Seed(time.Now().Unix())
+}
+
 type taskMaster struct {
 	initTime    time.Time
 	nextUpdate  time.Time
@@ -224,9 +229,11 @@ func (tm *taskMaster) schedule() (err error) {
 // Send retry failed tasks to tm.failedTopic (only if the phase exists in the workflow)
 func (tm *taskMaster) Process(t *task.Task) error {
 	meta, _ := url.ParseQuery(t.Meta)
-	// attempt to return
+
+	// attempt to retry
 	if t.Result == task.ErrResult {
 		p := tm.Get(*t)
+		rules, _ := url.ParseQuery(p.Rule)
 
 		r := meta.Get("retry")
 		i, _ := strconv.Atoi(r)
@@ -235,13 +242,23 @@ func (tm *taskMaster) Process(t *task.Task) error {
 			return nil
 		}
 		if p.Retry > i {
+			var delay time.Duration
+			if s := rules.Get("retry_delay"); s != "" {
+				delay, _ = time.ParseDuration(s)
+				delay = delay + jitterPercent(delay, 40)
+				meta.Set("delayed", gtools.PrintDuration(delay))
+			}
 			t = task.NewWithID(t.Type, t.Info, t.ID)
 			i++
 			meta.Set("retry", strconv.Itoa(i))
 			t.Meta = meta.Encode()
-			if err := tm.producer.Send(t.Type, t.JSONBytes()); err != nil {
-				return err
-			}
+			go func() {
+				time.Sleep(delay)
+				if err := tm.producer.Send(t.Type, t.JSONBytes()); err != nil {
+					log.Println(err)
+				}
+			}()
+			return nil
 		} else if tm.failedTopic != "-" {
 			// send to the retry failed topic if retries > p.Retry
 			meta.Set("retry", "failed")
@@ -321,3 +338,25 @@ func (tm *taskMaster) read(ctx context.Context) {
 		}
 	}
 }
+
+// jitterPercent will return a time.Duration representing extra
+// 'jitter' to be added to the wait time. Jitter is important
+// in retry events since the original cause of failure can be
+// due to too many jobs being processed at a time.
+//
+// By adding some jitter the retry events won't all happen
+// at once but will get staggered to prevent the problem
+// from happening again.
+//
+// 'p' is a percentage of the wait time. Duration returned
+// is a random duration between 0 and p. 'p' should be a value
+// between 0-100.
+func jitterPercent(wait time.Duration, p int64) time.Duration {
+	// p == 40
+	if wait == 0 {
+		return 0
+	}
+	maxJitter := (int64(wait) * p) / 100
+
+	return time.Duration(rand.Int63n(maxJitter))
+}
diff --git a/apps/taskmasters/flowlord/taskmaster_test.go b/apps/taskmasters/flowlord/taskmaster_test.go
index 2b77424f..e7671b08 100644
--- a/apps/taskmasters/flowlord/taskmaster_test.go
+++ b/apps/taskmasters/flowlord/taskmaster_test.go
@@ -2,8 +2,11 @@ package main
 
 import (
 	"encoding/json"
+	"regexp"
 	"sort"
+	"strings"
 	"testing"
+	"time"
 
 	"github.com/hydronica/trial"
 	"github.com/pcelvng/task"
@@ -13,6 +16,7 @@ import (
 )
 
 func TestTaskMaster_Process(t *testing.T) {
+	delayRegex := regexp.MustCompile(`delayed=(\d+.\d+)`)
 	cache, err := workflow.New("../../../internal/test/workflow/f1.toml", nil)
 	if err != nil {
 		t.Fatal("cache init", err)
@@ -31,6 +35,7 @@ func TestTaskMaster_Process(t *testing.T) {
 		tm.producer = producer
 		nop.FakeMsg = tsk.JSONBytes()
 		err = tm.Process(&tsk)
+		time.Sleep(100 * time.Millisecond)
 		result := make([]task.Task, 0)
 		for _, msgs := range producer.Messages {
 			for _, msg := range msgs {
@@ -39,6 +44,9 @@ func TestTaskMaster_Process(t *testing.T) {
 					return nil, err
 				}
 				v.Created = ""
+				if s := delayRegex.FindStringSubmatch(v.Meta); len(s) > 1 {
+					v.Meta = strings.Replace(v.Meta, s[1], "XX", 1)
+				}
 				result = append(result, v)
 			}
 		}
@@ -62,7 +70,7 @@ func TestTaskMaster_Process(t *testing.T) {
 					Type: "task1",
 					Info: "?date=2019-12-12",
 					ID:   "UUID_task1_attempt0",
-					Meta: "retry=1&workflow=f1.toml"},
+					Meta: "delayed=XXms&retry=1&workflow=f1.toml"},
 			},
 		},
 		"task1 attempt 2": {
@@ -77,7 +85,7 @@ func TestTaskMaster_Process(t *testing.T) {
 					Type: "task1",
 					Info: "?date=2019-12-12",
 					ID:   "UUID_task1_attempt2",
-					Meta: "retry=3&workflow=f1.toml"},
+					Meta: "delayed=XXms&retry=3&workflow=f1.toml"},
 			},
 		},
 		"task1 no retry": {
diff --git a/apps/workers/sql-load/worker.go b/apps/workers/sql-load/worker.go
index a3d308f3..9fa8b3c6 100644
--- a/apps/workers/sql-load/worker.go
+++ b/apps/workers/sql-load/worker.go
@@ -132,12 +132,12 @@ func (w *worker) DoTask(ctx context.Context) (task.Result, string) {
 		queryChan := make(chan string, 10)
 		go CreateInserts(rowChan, queryChan, tempTable, w.ds.insertCols, w.Params.BatchSize)
 
-		first := true
+		tableCreated := false
 		// load data into temp table
 		for s := range queryChan {
-			if first {
+			if !tableCreated {
 				s = createTempTable + s
-				first = false
+				tableCreated = true
 			}
 			if _, err := w.sqlDB.ExecContext(ctx, s); err != nil {
 				cancelFn()
@@ -149,6 +149,10 @@ func (w *worker) DoTask(ctx context.Context) (task.Result, string) {
 			return task.Failed(w.ds.err)
 		}
 
+		if !tableCreated {
+			return task.Completed("no data to load for %s", w.Params.Table)
+		}
+
 		//finalize and transfer data
 		var txErr error
 		var tx *sql.Tx
diff --git a/internal/test/workflow/f1.toml b/internal/test/workflow/f1.toml
index 096a8d6d..372ad13b 100644
--- a/internal/test/workflow/f1.toml
+++ b/internal/test/workflow/f1.toml
@@ -1,6 +1,6 @@
 [[Phase]]
 task = "task1"
-rule = "cron=0 * * * *&offset=-4h&job=t2"
+rule = "cron=0 * * * *&offset=-4h&job=t2&retry_delay=10ms"
 retry = 3
 template = "?date={yyyy}-{mm}-{dd}T{hh}"
 
@@ -33,7 +33,7 @@ template = "?date={yyyy}-{mm}-{dd}T{hh}"
 
 [[Phase]]
 task = "task5"
-rule = "job=t5"
+rule = "job=t5&retry_delay=10ms"
 retry = 3
 template = "?year={yyyy}"
 DependsOn = "task1:t2"
diff --git a/retry/retry.go b/retry/retry.go
index 03d4862d..aea8e272 100644
--- a/retry/retry.go
+++ b/retry/retry.go
@@ -229,7 +229,7 @@ func (r *Retryer) close() error {
 	return nil
 }
 
-// genJitter will return a time.Duration representing extra
+// jitterPercent will return a time.Duration representing extra
 // 'jitter' to be added to the wait time. Jitter is important
 // in retry events since the original cause of failure can be
 // due to too many jobs being processed at a time.
diff --git a/workflow/workflow.go b/workflow/workflow.go
index 8daa3179..55ef9951 100644
--- a/workflow/workflow.go
+++ b/workflow/workflow.go
@@ -285,12 +285,14 @@ func (c *Cache) loadFile(path string, opts *file.Options) (f string, err error)
 	if err != nil {
 		return "", errors.Wrapf(err, "read-all: %s", path)
 	}
-
-	if _, err := toml.Decode(string(b), &data); err != nil {
+	d := Workflow{
+		Checksum: data.Checksum,
+	}
+	if _, err := toml.Decode(string(b), &d); err != nil {
 		return "", errors.Wrapf(err, "decode: %s", string(b))
 	}
 
-	c.Workflows[f] = data
+	c.Workflows[f] = d
 
 	return f, nil
 }
diff --git a/workflow/workflow_test.go b/workflow/workflow_test.go
index 1a092ab0..b3132a2c 100644
--- a/workflow/workflow_test.go
+++ b/workflow/workflow_test.go
@@ -26,7 +26,7 @@ func TestLoadFile(t *testing.T) {
 	cases := trial.Cases{
 		"read file": {
 			Input:    input{path: "../internal/test/workflow/f1.toml"},
-			Expected: "71c745e929d87c1448e516877acbc86f", // checksum of test file
+			Expected: "4422274d9c9f7e987c609687a7702651", // checksum of test file
 		},
 		"stat error": {
 			Input:       input{path: "nop://stat_err"},