我已使用flatfile选项手动启动了具有3个节点的H2O群集。
然后,我尝试使用以下方式连接到该群集:
tuple
在我的3个节点的H2O日志中,我看到:
from pysparkling import *
conf = H2OConf(spark)
.set_external_cluster_mode()
.use_manual_cluster_start()
.set_h2o_cluster("my ip", 54321)
.set_cloud_name("ClusterName")
hc = H2OContext.getOrCreate(spark, conf)
显然会导致pyspark的另一个错误:
12-19 13:29:09.717 X.X.X.X:54321 5597 FJ-126-7 INFO: Client reported via broadcast message /X.X.X.X:54321 from ip-X-X-X-X.ec2.internal/X.X.X.X:54321
12-19 13:29:09.717 X.X.X.X:54321 5597 FJ-126-7 INFO: New client discovered: /X.X.X.X:54321(watchdog=false, cloud_name_hash=125572707)
12-19 13:29:12.854 X.X.X.X:54321 5597 #ckThread WARN: Client /X.X.X.X:54321 disconnected!
我不知道断开连接来自哪里... 有人遇到过类似的问题吗?
非常感谢!