From 2feba4409c447ee01bba269a159c44b2b6b741db Mon Sep 17 00:00:00 2001 From: Edward Oakes Date: Mon, 9 Nov 2020 11:21:03 -0600 Subject: [PATCH] [serve] Fix long running failure test (#11805) --- .../workloads/serve_failure.py | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/release/long_running_tests/workloads/serve_failure.py b/release/long_running_tests/workloads/serve_failure.py index 9f5db466e..129853289 100644 --- a/release/long_running_tests/workloads/serve_failure.py +++ b/release/long_running_tests/workloads/serve_failure.py @@ -32,8 +32,8 @@ client = serve.start(detached=True) @ray.remote class RandomKiller: - def __init__(self, client, kill_period_s=1): - self.client = client + def __init__(self, kill_period_s=1): + self.client = serve.connect() self.kill_period_s = kill_period_s def _get_all_serve_actors(self): @@ -55,8 +55,8 @@ class RandomKiller: class RandomTest: - def __init__(self, client, max_endpoints=1): - self.client = client + def __init__(self, serve_client, max_endpoints=1): + self.client = serve_client self.max_endpoints = max_endpoints self.weighted_actions = [ (self.create_endpoint, 1), @@ -114,8 +114,8 @@ class RandomTest: iteration += 1 -random_killer = RandomKiller.remote(client) +random_killer = RandomKiller.remote() random_killer.run.remote() -# Subtract 4 from the CPUs available for master, router, HTTP proxy, -# and metric monitor actors. -RandomTest(client, max_endpoints=(num_nodes * cpus_per_node) - 4).run() +# Subtract 1 CPU for the controller and 1 CPU from each node for the HTTP +# server. +RandomTest(client, max_endpoints=(num_nodes * (cpus_per_node - 1)) - 1).run()