前提条件

1. 已创建Redis实例且已设置密码，如未创建请参考创建实例。
2. 已创建EMR Ray集群，创建Ray集群请参考创建虚拟集群。
3. EMR Ray 集群与 Redis 处于相同网络且 Ray 集群所在网络在 Redis 实例的白名单中。
4. Redis 实例中的disabled-commands参数需移除 keys 以允许执行 keys 命令。

创建 GCS HA Ray cluster

1. 在 服务列表中单击 Ray 进入部署拓扑页面。

2. 点击添加RayCluster，在添加RayCluster界面打开GCS高可用，并选择对应的Redis实例，填写对应的Redis密码，点击**确定，**创建 RayCluster。

验证

在 Ray Head Pod 中创建两个 python 文件

detached_actor.py

import ray

@ray.remote(num_cpus=1)
class Counter:
    def __init__(self):
        self.value = 0

    def increment(self):
        self.value += 1
        return self.value

ray.init(namespace="default_namespace")
Counter.options(name="counter_actor", lifetime="detached").remote()

increment_counter.py

import ray

ray.init(namespace="default_namespace")
counter = ray.get_actor("counter_actor")
print(ray.get(counter.increment.remote()))

创建 actor

python3 detached_actor.py
2025-01-21 21:48:55,351 INFO worker.py:1461 -- Using address 127.0.0.1:6379 set in the environment variable RAY_ADDRESS
2025-01-21 21:48:55,351 INFO worker.py:1601 -- Connecting to existing Ray cluster at address: 10.1.0.134:6379...
2025-01-21 21:48:55,357 INFO worker.py:1777 -- Connected to Ray cluster. View the dashboard at http://10.1.0.134:8265 

python3 increment_counter.py

2025-01-21 21:49:18,140 INFO worker.py:1461 -- Using address 127.0.0.1:6379 set in the environment variable RAY_ADDRESS
2025-01-21 21:49:18,141 INFO worker.py:1601 -- Connecting to existing Ray cluster at address: 10.1.0.134:6379...
2025-01-21 21:49:18,146 INFO worker.py:1777 -- Connected to Ray cluster. View the dashboard at http://10.1.0.134:8265 
1

终止 GCS 服务

ps aux | grep gcs_server | grep -v grep | awk '{print $2}' | xargs kill -9

head 恢复后查看 actor 值

返回值为2，符合预期。

python3 increment_counter.py
2025-01-21 21:53:12,456 INFO worker.py:1461 -- Using address 127.0.0.1:6379 set in the environment variable RAY_ADDRESS
2025-01-21 21:53:12,456 INFO worker.py:1601 -- Connecting to existing Ray cluster at address: 10.1.0.134:6379...
2025-01-21 21:53:12,462 INFO worker.py:1777 -- Connected to Ray cluster. View the dashboard at http://10.1.0.134:8265 
2