From 9d0148dbbe418bcf4a4456e5b029ae721ee5fba3 Mon Sep 17 00:00:00 2001
From: SangBin Cho <rkooo567@gmail.com>
Date: Mon, 7 Mar 2022 11:24:54 +0900
Subject: [PATCH] [Test] Migrate the first test to the new infra (#22770)

This migrate the simplest nightly test to the new infra. I will also explore k8s migration with this test
---
 release/nightly_tests/nightly_tests.yaml      | 16 ++++++-------
 release/nightly_tests/wait_cluster.py         | 11 +--------
 .../command_runner/_wait_cluster.py           |  1 +
 release/release_tests.yaml                    | 23 +++++++++++++++++++
 4 files changed, 33 insertions(+), 18 deletions(-)

diff --git a/release/nightly_tests/nightly_tests.yaml b/release/nightly_tests/nightly_tests.yaml
index 8c20382f9..1836ae07e 100644
--- a/release/nightly_tests/nightly_tests.yaml
+++ b/release/nightly_tests/nightly_tests.yaml
@@ -3,15 +3,15 @@
 #
 # Test basic single node 10GB shuffle with a small number of partitions.
 # This doesn't require object spilling.
-- name: shuffle_10gb
-  team: core
-  cluster:
-    app_config: shuffle/shuffle_app_config.yaml
-    compute_template: shuffle/shuffle_compute_single.yaml
+# - name: shuffle_10gb
+#   team: core
+#   cluster:
+#     app_config: shuffle/shuffle_app_config.yaml
+#     compute_template: shuffle/shuffle_compute_single.yaml
 
-  run:
-    timeout: 3000
-    script: python shuffle/shuffle_test.py --num-partitions=50 --partition-size=200e6
+#   run:
+#     timeout: 3000
+#     script: python shuffle/shuffle_test.py --num-partitions=50 --partition-size=200e6
 
 # Test single node 50GB shuffle with a large number of partitions.
 - name: shuffle_50gb
diff --git a/release/nightly_tests/wait_cluster.py b/release/nightly_tests/wait_cluster.py
index cbabb9a5e..f70088289 100644
--- a/release/nightly_tests/wait_cluster.py
+++ b/release/nightly_tests/wait_cluster.py
@@ -26,15 +26,6 @@ start = time.time()
 next_feedback = start
 max_time = start + args.max_time_s
 
-
-def num_alive_nodes():
-    n = 0
-    for node in ray.nodes():
-        if node.get("Alive", False):
-            n += 1
-    return n
-
-
 while not curr_nodes >= args.num_nodes:
     now = time.time()
 
@@ -54,7 +45,7 @@ while not curr_nodes >= args.num_nodes:
         next_feedback = now + args.feedback_interval_s
 
     time.sleep(5)
-    curr_nodes = num_alive_nodes()
+    curr_nodes = len(ray.nodes())
 
 passed = time.time() - start
 print(
diff --git a/release/ray_release/command_runner/_wait_cluster.py b/release/ray_release/command_runner/_wait_cluster.py
index c02330db2..f70088289 100644
--- a/release/ray_release/command_runner/_wait_cluster.py
+++ b/release/ray_release/command_runner/_wait_cluster.py
@@ -25,6 +25,7 @@ curr_nodes = 0
 start = time.time()
 next_feedback = start
 max_time = start + args.max_time_s
+
 while not curr_nodes >= args.num_nodes:
     now = time.time()
 
diff --git a/release/release_tests.yaml b/release/release_tests.yaml
index cc94daac2..96328970b 100644
--- a/release/release_tests.yaml
+++ b/release/release_tests.yaml
@@ -761,3 +761,26 @@
 
   alert: tune_tests
 
+########################
+# Core Nightly Tests
+########################
+
+- name: shuffle_10gb
+  group: Core nightly tests
+  team: core
+  working_dir: nightly_tests
+
+  legacy:
+    test_name: shuffle_10gb
+    test_suite: nightly_tests
+
+  cluster:
+    cluster_env: shuffle/shuffle_app_config.yaml
+    cluster_compute: shuffle/shuffle_compute_single.yaml
+
+  run:
+    timeout: 3000
+    script: python shuffle/shuffle_test.py --num-partitions=50 --partition-size=200e6
+    
+    type: sdk_command
+    file_manager: sdk