From ab8cfb9f57eba4c53290eb7533bd375c4aa95b82 Mon Sep 17 00:00:00 2001
From: Sam Kleinman <garen@tychoish.com>
Date: Fri, 24 Sep 2021 12:28:51 -0400
Subject: [PATCH] e2e: tighten timing for load generation (#6990)

---
 .github/workflows/e2e-nightly-master.yml |  4 ++--
 test/e2e/runner/load.go                  | 24 +++++++++---------------
 test/e2e/runner/main.go                  | 16 +++++++++++++++-
 3 files changed, 26 insertions(+), 18 deletions(-)

diff --git a/.github/workflows/e2e-nightly-master.yml b/.github/workflows/e2e-nightly-master.yml
index 029fee6bb..5f5ed9706 100644
--- a/.github/workflows/e2e-nightly-master.yml
+++ b/.github/workflows/e2e-nightly-master.yml
@@ -17,7 +17,7 @@ jobs:
       fail-fast: false
       matrix:
         p2p: ['legacy', 'new', 'hybrid']
-        group: ['00', '01']
+        group: ['00', '01', '02', '03']
     runs-on: ubuntu-latest
     timeout-minutes: 60
     steps:
@@ -35,7 +35,7 @@ jobs:
       - name: Generate testnets
         working-directory: test/e2e
         # When changing -g, also change the matrix groups above
-        run: ./build/generator -g 2 -d networks/nightly/${{ matrix.p2p }} -p ${{ matrix.p2p }}
+        run: ./build/generator -g 4 -d networks/nightly/${{ matrix.p2p }} -p ${{ matrix.p2p }}
 
       - name: Run ${{ matrix.p2p }} p2p testnets in group ${{ matrix.group }}
         working-directory: test/e2e
diff --git a/test/e2e/runner/load.go b/test/e2e/runner/load.go
index d66d7770e..5e6e04f89 100644
--- a/test/e2e/runner/load.go
+++ b/test/e2e/runner/load.go
@@ -3,7 +3,6 @@ package main
 import (
 	"container/ring"
 	"context"
-	"errors"
 	"fmt"
 	"math/rand"
 	"time"
@@ -58,19 +57,9 @@ func Load(ctx context.Context, testnet *e2e.Testnet) error {
 		case numSeen := <-chSuccess:
 			success += numSeen
 		case <-ctx.Done():
-			// if we couldn't submit any transactions,
-			// that's probably a problem and the test
-			// should error; however, for very short tests
-			// we shouldn't abort.
-			//
-			// The 2s cut off, is a rough guess based on
-			// the expected value of
-			// loadGenerateWaitTime. If the implementation
-			// of that function changes, then this might
-			// also need to change without more
-			// refactoring.
-			if success == 0 && time.Since(started) > 2*time.Second {
-				return errors.New("failed to submit any transactions")
+			if success == 0 {
+				return fmt.Errorf("failed to submit transactions in %s by %d workers",
+					time.Since(started), concurrency)
 			}
 
 			// TODO perhaps allow test networks to
@@ -141,9 +130,14 @@ func loadGenerateWaitTime(size int64) time.Duration {
 		baseJitter = rand.Int63n(max-min+1) + min // nolint: gosec
 		sizeFactor = size * int64(time.Millisecond)
 		sizeJitter = rand.Int63n(sizeFactor-min+1) + min // nolint: gosec
+		waitTime   = time.Duration(baseJitter + sizeJitter)
 	)
 
-	return time.Duration(baseJitter + sizeJitter)
+	if size == 1 {
+		return waitTime / 2
+	}
+
+	return waitTime
 }
 
 // loadProcess processes transactions
diff --git a/test/e2e/runner/main.go b/test/e2e/runner/main.go
index c6d2ad7b7..194e4c6e4 100644
--- a/test/e2e/runner/main.go
+++ b/test/e2e/runner/main.go
@@ -5,6 +5,7 @@ import (
 	"fmt"
 	"os"
 	"strconv"
+	"time"
 
 	"github.com/spf13/cobra"
 
@@ -75,7 +76,7 @@ func NewCLI() *CLI {
 			go func() {
 				chLoadResult <- Load(lctx, cli.testnet)
 			}()
-
+			startAt := time.Now()
 			if err = Start(ctx, cli.testnet); err != nil {
 				return err
 			}
@@ -102,6 +103,19 @@ func NewCLI() *CLI {
 				}
 			}
 
+			// to help make sure that we don't run into
+			// situations where 0 transactions have
+			// happened on quick cases, we make sure that
+			// it's been at least 10s before canceling the
+			// load generator.
+			//
+			// TODO allow the load generator to report
+			// successful transactions to avoid needing
+			// this sleep.
+			if rest := time.Since(startAt); rest < 15*time.Second {
+				time.Sleep(15*time.Second - rest)
+			}
+
 			loadCancel()
 
 			if err = <-chLoadResult; err != nil {