hiro/ray
1
0
Fork 0
mirror of https://github.com/vale981/ray synced 2025-03-06 10:31:39 -05:00
Code Issues Projects Releases Packages Wiki Activity Actions

[autoscaler] Interpret autoscaling_speed as 1/x-1 of previous target util fraction (#11961)

Browse source 
* tweak

* update
This commit is contained in:
Eric Liang 2020-11-12 16:23:50 -08:00 • committed by GitHub
parent 0bd69edd71
commit a6a8e777f3
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 8 additions and 8 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

5
python/ray/autoscaler/_private/autoscaler.py
View file

@ -362,11 +362,12 @@ class StandardAutoscaler:
"detected. Replacing it by setting upscaling_speed to "
"99999.")
elif target_utilization_fraction:
upscaling_speed = 1 / max(target_utilization_fraction, 0.001)
upscaling_speed = (
1 / max(target_utilization_fraction, 0.001) - 1)
logger.warning(
"Legacy target_utilization_fraction config "
"detected. Replacing it by setting upscaling_speed to " +
"1 / target_utilization_fraction.")
"1 / target_utilization_fraction - 1.")
else:
upscaling_speed = 1.0
if self.resource_demand_scheduler:

11
python/ray/tests/test_autoscaling_policy.py
View file

@ -415,17 +415,16 @@ class Simulator:
num_connected_nodes = len(
[node for node in self.ip_to_nodes.values() if node.in_cluster])
num_pending_nodes = len(self.ip_to_nodes) - num_connected_nodes
return f"""[t={self.virtual_time}]
Connected nodes: {num_connected_nodes}
Pending nodes: {num_pending_nodes}
Remaining requests: {len(self.work_queue)}
"""
return (f"[t={self.virtual_time}] "
f"Connected: {num_connected_nodes}, "
f"Pending: {num_pending_nodes}, "
f"Remaining: {len(self.work_queue)}")
SAMPLE_CLUSTER_CONFIG = copy.deepcopy(MULTI_WORKER_CLUSTER)
SAMPLE_CLUSTER_CONFIG["min_workers"] = 0
SAMPLE_CLUSTER_CONFIG["max_workers"] = 9999
SAMPLE_CLUSTER_CONFIG["target_utilization_fraction"] = 1.0
SAMPLE_CLUSTER_CONFIG["target_utilization_fraction"] = 0.5
SAMPLE_CLUSTER_CONFIG["available_node_types"]["m4.16xlarge"][
"max_workers"] = 100
SAMPLE_CLUSTER_CONFIG["available_node_types"]["m4.4xlarge"][