Rstudio 服务器：无法在 Centos 上启动 H2O.init

Question

H2o 无法成功启动，但我仍然可以访问链接 本地主机：12345/flow/index.html

登录 R Studio

h2o.init 中的错误（nthreads = -1，max_mem_size = “4g”，ip = “localhost”，： H2O 启动失败，停止执行。

登录操作系统

[root@xxxx]# tail -200f /tmp/RtmpYou3gq/h2o_tuongnv4_started_from_r.out


: ----- H2O started  -----
: Build git branch: rel-ueno
: Build git hash: f521cfeb6cf6c41ca6a728cdbaafe96509170907
: Build git describe: jenkins-3.10.4.5-4-gf521cfe
: Build project version: 3.10.4.6 (latest version: unknown)
: Build age: 15 days
: Built by: 'jenkins'
: Built on: '2017-04-26 22:11:30'
: Watchdog Build git branch: (unknown)
: Watchdog Build git hash: (unknown)
: Watchdog Build git describe: (unknown)
: Watchdog Build project version: (unknown)
: Watchdog Built by: (unknown)
: Watchdog Built on: (unknown)
: Processed H2O arguments: [-name, H2O_started_from_R_tuongnv4_tju329, -ip, localhost, -port, 12345, -ice_root, /tmp/RtmpYou3gq]
: Java availableProcessors: 8
: Java heap totalMemory: 238.0 MB
: Java heap maxMemory: 3.56 GB
: Java version: Java 1.8.0_131 (from Oracle Corporation)
: JVM launch parameters: [-Xmx4g, -ea]
: OS version: Linux 3.10.0-514.el7.x86_64 (amd64)
: Machine physical memory: 15.44 GB
: X-h2o-cluster-id: 1494587541855
: User name: 'tuongnv4'
: Opted out of sending usage metrics.
: IPv6 stack selected: false
: Possible IP Address: enp0s31f6 (enp0s31f6), 192.168.170.166
: Possible IP Address: lo (lo), 127.0.0.1
: Selected H2O.CLOUD_MULTICAST_IF: name:lo (lo) doesn't support multicast
: H2O node running in unencrypted mode.
: Internal communication uses port: 12346
: Listening for HTTP and REST traffic on http://127.0.0.1:12345/
: H2O cloud name: 'H2O_started_from_R_tuongnv4_tju329' on localhost/127.0.0.1:12345, discovery address /228.246.241.107:58614
: If you have trouble connecting, try SSH tunneling from your local machine (e.g., via port 55555):
:   1. Open a terminal and run 'ssh -L 55555:localhost:12345 tuongnv4@127.0.0.1'
:   2. Point your browser to http://localhost:55555
: Log dir: '/tmp/RtmpYou3gq/h2ologs'
: Cur dir: '/home/tuongnv4'
: HDFS subsystem successfully initialized
: S3 subsystem successfully initialized
: Flow dir: '/home/tuongnv4/h2oflows'
: Cloud of size 1 formed [localhost/127.0.0.1:12345]
: Registered parsers: [GUESS, ARFF, XLS, SVMLight, AVRO, PARQUET, CSV]
: Watchdog extension initialized
: Registered 1 extensions in: 2mS
: Registered: 148 REST APIs in: 283mS
: Registered: 209 schemas in 333ms
: H2O started in 2128ms
: 
: Open H2O Flow in your web browser: http://127.0.0.1:12345
: 
: GET /flow/index.html, parms: {}
: GET /flow/index.html, parms: {}
: GET /3/Metadata/endpoints, parms: {}
: Locking cloud to new members, because hex.schemas.DeepLearningV3
: GET /flow/help/catalog.json, parms: {}
: GET /3/ModelBuilders, parms: {}
: GET /3/NodePersistentStorage/categories/environment/names/clips/exists, parms: {}
: GET /3/About, parms: {}
: GET /3/NodePersistentStorage/notebook, parms: {}

Answer 1

几乎可以肯定这里发生的是服务器配置问题。

对于初学者，我建议：

• 禁用任何 linux 防火墙 (iptables/iptables6)
• 禁用 SElinux

然后再试一次。

Answer 2

您从非标准端口（12345 而不是默认的 54321）开始。

尝试将port 参数删除到h2o.init()，它可能会正常工作。

（在处理端口参数时存在或曾经存在错误，这可能是您遇到的问题。我相信解决方法是从命令行手动启动 h2o.jar（即在 R 之外） ，然后h2o.init(port=12345) 应该能够连接到它。）

Answer 3

我能想到另一件事可能会发生......

如果您能够使用 Web 浏览器连接到 H2O，但不能从 R 中连接，那么您可能需要从 R 中取消设置代理环境变量。

确保在 R 会话中未设置环境变量 http_proxy 和 https_proxy。