fix

Signed-off-by: Ryan Leung <rleungx@gmail.com>

tests/integrations/realcluster/Makefile

@@ -28,10 +28,8 @@ tidy:
 	git diff go.mod go.sum | cat
 	git diff --quiet go.mod go.sum
 
-.PHONY: check
 check: tiup test
 
-.PHONY: tiup
 tiup:
 	@echo "==> Checking tiup installation"
 	@if ! which tiup > /dev/null 2>&1; then \
@@ -42,71 +40,38 @@ tiup:
 	fi
 	@echo "tiup version: $$(tiup --version)"
 
-.PHONY: deploy
 deploy: kill_cluster deploy_only
 
-.PHONY: deploy_only
 deploy_only:
-	@echo "==> Deploying cluster..."
-	@./deploy.sh
-	@echo "==> Waiting for cluster to be ready..."
-	@./wait_tiup.sh 15 20
-	@echo "==> Checking cluster status..."
-	@if pid=$$(pgrep -f "tiup.*playground.*--tag"); then \
-		echo "Playground process running with pid: $$pid"; \
-		tiup playground display; \
-	else \
-		echo "ERROR: Playground process not found"; \
-		exit 1; \
-	fi
+	@ echo "deploying..."
+	./deploy.sh
+	@ echo "wait cluster ready..."
+	./wait_tiup.sh 15 20
+	@ echo "check cluster status..."
+	@ pid=$$(ps -ef | grep 'playground' | grep -v grep | awk '{print $$2}' | head -n 1); \
+	echo $$pid;
 
-.PHONY: kill_cluster
 kill_cluster:
-	@echo "==> Killing cluster..."
-	@if pid=$$(pgrep -f "tiup.*playground.*--tag"); then \
-		echo "Found playground process(es):"; \
-		ps -fp $$pid; \
-		echo "Sending SIGTERM..."; \
+	@ echo "kill cluster..."
+	@ pid=$$(ps -ef | grep 'playground' | grep -v grep | awk '{print $$2}' | head -n 1); \
+	if [ ! -z "$$pid" ]; then \
+		echo $$pid; \
 		kill $$pid; \
-		echo "Waiting for process to exit (30s)..."; \
-		for i in $$(seq 1 30); do \
-			if ! kill -0 $$pid 2>/dev/null; then \
-				echo "Process exited"; \
-				exit 0; \
-			fi; \
-			sleep 1; \
-		done; \
-		echo "Process still running, sending SIGKILL..."; \
-		kill -9 $$pid || true; \
-	else \
-		echo "No playground process found"; \
+		echo "waiting for cluster to exit..."; \
+		sleep 30; \
 	fi
 
-.PHONY: test
 test:
-	@echo "==> Running integration tests..."
-	@CGO_ENABLED=1 go test ./... -v -tags deadlock -race -cover || (\
-		echo "==> Test failed. Collecting logs..."; \
-		$(MAKE) collect_logs; \
-		exit 1 \
-	)
-
-.PHONY: collect_logs
-collect_logs:
-	@echo "==> Collecting cluster logs..."
-	@for pd in 0 1 2; do \
-		echo "### PD-$$pd logs ###"; \
-		cat ~/.tiup/data/pd_real_cluster_test/pd-$$pd/pd.log 2>/dev/null || echo "No log file found"; \
-		echo; \
-	done
-	@echo "==> Collecting playground logs..."
-	@find /tmp/real_cluster/playground -type f -name "*.log" -exec sh -c 'echo "### $$(basename {}) ###"; cat {}; echo' \;
-
-.PHONY: clean
-clean:
-	@echo "==> Cleaning up..."
-	@rm -rf /tmp/real_cluster/playground/*
-	@echo "==> Cleaned playground logs"
+	CGO_ENABLED=1 go test ./... -v -tags deadlock -race -cover || (\
+	echo "The following is pd-0 log" ; \
+	cat ~/.tiup/data/pd_real_cluster_test/pd-0/pd.log ; \
+	echo "The following is pd-1 log" ; \
+	cat ~/.tiup/data/pd_real_cluster_test/pd-1/pd.log ; \
+	echo "The following is pd-2 log" ; \
+	cat ~/.tiup/data/pd_real_cluster_test/pd-2/pd.log ; \
+	exit 1)
 
 install-tools:
 	cd $(ROOT_PATH) && $(MAKE) install-tools
+
+.PHONY: tidy static check tiup deploy deploy_only kill_cluster test install-tools

tests/integrations/realcluster/cluster.go

@@ -143,12 +143,13 @@ func (c *cluster) stop() {
 			_ = syscall.Kill(pid, syscall.SIGKILL)
 		}
 	}
-	log.Info("cluster destroyed", zap.String("tag", c.tag))
+	log.Info("cluster stopped", zap.String("tag", c.tag))
 }
 
 func (c *cluster) deploy() {
 	re := c.re
 	curPath, err := os.Getwd()
+	log.Info(curPath)
 	re.NoError(err)
 	re.NoError(os.Chdir("../../.."))
 
@@ -218,18 +219,13 @@ func (c *cluster) collectPids() error {
 
 func (c *cluster) waitReady() {
 	re := c.re
-	const (
-		interval = 5
-		maxTimes = 20
-	)
 	log.Info("start to wait TiUP ready", zap.String("tag", c.tag))
-	timeout := time.After(time.Duration(maxTimes*interval) * time.Second)
-	ticker := time.NewTicker(time.Duration(interval) * time.Second)
+	ticker := time.NewTicker(5 * time.Second)
 	defer ticker.Stop()
 
-	for i := range maxTimes {
+	for {
 		select {
-		case <-timeout:
+		case <-time.After(100 * time.Second):
 			re.FailNowf("TiUP is not ready after timeout, tag: %s", c.tag)
 		case <-ticker.C:
 			log.Info("check TiUP ready", zap.String("tag", c.tag))
@@ -240,11 +236,9 @@ func (c *cluster) waitReady() {
 				return
 			}
 			log.Info(output)
-			log.Info("TiUP is not ready, will retry", zap.Int("retry times", i),
-				zap.String("tag", c.tag), zap.Error(err))
+			log.Info("TiUP is not ready, will retry", zap.String("tag", c.tag), zap.Error(err))
 		}
 	}
-	re.FailNowf("TiUP is not ready after max retries, tag: %s", c.tag)
 }
 
 func buildBinPathsOpts(ms bool) string {