Change logfile names and also allow plasma store socket to be passed in. (#2862)

test/stress_tests.py

@@ -8,6 +8,7 @@ import pytest
 import time
 
 import ray
+import ray.tempfile_services
 import ray.ray_constants as ray_constants
 
 
@@ -173,10 +174,10 @@ def ray_start_reconstruction(request):
     plasma_addresses = []
     objstore_memory = plasma_store_memory // num_local_schedulers
     for i in range(num_local_schedulers):
-        store_stdout_file, store_stderr_file = ray.services.new_log_files(
-            "plasma_store_{}".format(i), True)
-        manager_stdout_file, manager_stderr_file = (ray.services.new_log_files(
-            "plasma_manager_{}".format(i), True))
+        store_stdout_file, store_stderr_file = (
+            ray.tempfile_services.new_plasma_store_log_file(i, True))
+        manager_stdout_file, manager_stderr_file = (
+            ray.tempfile_services.new_plasma_manager_log_file(i, True))
         plasma_addresses.append(
             ray.services.start_plasma_store(
                 node_ip_address,

test/tempfile_test.py

@@ -0,0 +1,119 @@
+import os
+import shutil
+import time
+import pytest
+import ray
+import ray.tempfile_services as tempfile_services
+
+
+def test_conn_cluster():
+    # plasma_store_socket_name
+    with pytest.raises(Exception) as exc_info:
+        ray.init(
+            use_raylet=True,
+            redis_address="127.0.0.1:6379",
+            plasma_store_socket_name="/tmp/this_should_fail")
+    assert exc_info.value.args[0] == (
+        "When connecting to an existing cluster, "
+        "plasma_store_socket_name must not be provided.")
+
+    # raylet_socket_name
+    with pytest.raises(Exception) as exc_info:
+        ray.init(
+            use_raylet=True,
+            redis_address="127.0.0.1:6379",
+            raylet_socket_name="/tmp/this_should_fail")
+    assert exc_info.value.args[0] == (
+        "When connecting to an existing cluster, "
+        "raylet_socket_name must not be provided.")
+
+    # temp_dir
+    with pytest.raises(Exception) as exc_info:
+        ray.init(
+            use_raylet=True,
+            redis_address="127.0.0.1:6379",
+            temp_dir="/tmp/this_should_fail")
+    assert exc_info.value.args[0] == (
+        "When connecting to an existing cluster, "
+        "temp_dir must not be provided.")
+
+
+def test_tempdir():
+    ray.init(use_raylet=True, temp_dir="/tmp/i_am_a_temp_dir")
+    assert os.path.exists(
+        "/tmp/i_am_a_temp_dir"), "Specified temp dir not found."
+    ray.shutdown()
+    shutil.rmtree("/tmp/i_am_a_temp_dir", ignore_errors=True)
+
+
+def test_raylet_socket_name():
+    ray.init(use_raylet=True, raylet_socket_name="/tmp/i_am_a_temp_socket")
+    assert os.path.exists(
+        "/tmp/i_am_a_temp_socket"), "Specified socket path not found."
+    ray.shutdown()
+    try:
+        os.remove("/tmp/i_am_a_temp_socket")
+    except Exception:
+        pass
+
+
+def test_temp_plasma_store_socket():
+    ray.init(
+        use_raylet=True, plasma_store_socket_name="/tmp/i_am_a_temp_socket")
+    assert os.path.exists(
+        "/tmp/i_am_a_temp_socket"), "Specified socket path not found."
+    ray.shutdown()
+    try:
+        os.remove("/tmp/i_am_a_temp_socket")
+    except Exception:
+        pass
+
+
+def test_raylet_tempfiles():
+    ray.init(use_raylet=True, redirect_worker_output=False)
+    top_levels = set(os.listdir(tempfile_services.get_temp_root()))
+    assert top_levels == {"ray_ui.ipynb", "sockets", "logs"}
+    log_files = set(os.listdir(tempfile_services.get_logs_dir_path()))
+    assert log_files == {
+        "log_monitor.out", "log_monitor.err", "plasma_store_0.out",
+        "plasma_store_0.err", "webui.out", "webui.err", "monitor.out",
+        "monitor.err", "redis-shard_0.out", "redis-shard_0.err", "redis.out",
+        "redis.err"
+    }  # without raylet logs
+    socket_files = set(os.listdir(tempfile_services.get_sockets_dir_path()))
+    assert socket_files == {"plasma_store", "raylet"}
+    ray.shutdown()
+
+    ray.init(use_raylet=True, redirect_worker_output=True, num_workers=0)
+    top_levels = set(os.listdir(tempfile_services.get_temp_root()))
+    assert top_levels == {"ray_ui.ipynb", "sockets", "logs"}
+    log_files = set(os.listdir(tempfile_services.get_logs_dir_path()))
+    assert log_files == {
+        "log_monitor.out", "log_monitor.err", "plasma_store_0.out",
+        "plasma_store_0.err", "webui.out", "webui.err", "monitor.out",
+        "monitor.err", "redis-shard_0.out", "redis-shard_0.err", "redis.out",
+        "redis.err", "raylet_0.out", "raylet_0.err"
+    }  # with raylet logs
+    socket_files = set(os.listdir(tempfile_services.get_sockets_dir_path()))
+    assert socket_files == {"plasma_store", "raylet"}
+    ray.shutdown()
+
+    ray.init(use_raylet=True, redirect_worker_output=True, num_workers=2)
+    top_levels = set(os.listdir(tempfile_services.get_temp_root()))
+    assert top_levels == {"ray_ui.ipynb", "sockets", "logs"}
+    time.sleep(3)  # wait workers to start
+    log_files = set(os.listdir(tempfile_services.get_logs_dir_path()))
+    assert log_files.issuperset({
+        "log_monitor.out", "log_monitor.err", "plasma_store_0.out",
+        "plasma_store_0.err", "webui.out", "webui.err", "monitor.out",
+        "monitor.err", "redis-shard_0.out", "redis-shard_0.err", "redis.out",
+        "redis.err", "raylet_0.out", "raylet_0.err"
+    })  # with raylet logs
+
+    # Check numbers of worker log file.
+    assert sum(
+        1 for filename in log_files if filename.startswith("worker")) == 4
+
+    socket_files = set(os.listdir(tempfile_services.get_sockets_dir_path()))
+    assert socket_files == {"plasma_store", "raylet"}
+    ray.shutdown()