hi, Vinodkone,
When we had the line "127.0.1.1 hdfs2" in the file /etc/hosts, a Spark
standalone cluster worked correctly and the Spark streaming example
NetworkWordCount worked well on a standalone Spark cluster with one
master and two slaves. But the Mesos cluster didn't.
Xiaoying
于 2013/4/18 13:52, Vinod Kone 写道:
I see. Looks like Mesos components cannot talk from a non-local ip address
(e.g. 192.x.x.x) to local address (127.0.0.1), even though the reverse
direction works. I'm not sure if its a networking issue or if its unique to
how Mesos does communication between 2 endpoints.
The fact that you either had to manually specify --ip or change /etc/hosts,
tells me that by default Mesos picks host ip from /etc/hosts.
@vinodkone
On Wed, Apr 17, 2013 at 10:40 PM, Xiaoying Zheng <[email protected]> wrote:
hi,
We had a line "127.0.1.1 hdfs2" in /etc/hosts of the slave node, where
hdfs2 is the machine name of the slave. After removing the line, it is
alright now.
于 2013/4/18 13:24, Vinod Kone 写道:
Hi Xiaoying,
Could you tell us what you mean by fixing /etc/hosts?
Thanks,
On Wed, Apr 17, 2013 at 10:14 PM, Xiaoying Zheng <[email protected]>
wrote:
hi, Ben,
We fixed errors in our /etc/hosts configuration. It works well now.
Thanks.
Xiaoying
--
郑小盈 博士 副研究员
中国科学院上海高等研究院
高性能计算与数据中心
上海浦东新区张江高科技园区海科路99号 201210
邮箱: [email protected]
电话: 0086-21-2032-5118
Xiaoying Zheng, Ph.D., Associate Professor
R&D Center for Cloud Computing
Shanghai Advanced Research Institute, Chinese Academy of Sciences
Email: [email protected]
Tel.: 0086-21-2032-5118
