Unable to open H2O Flow in Databricks

[Ethan Zou]

Hi,

I am using Databricks community edition and I have installed Sparking Water on my spark cluster. Based on online instructions, I would like to access H2O's web UI called H2O Flow. But I couldn't open Flow through command "openFlow". Can anyone please help? Attached please find my screenshot.

Thanks,

Ethan

[Lauren DiPerna]

Hi Ethan,

Were you able to open Flow by entering the url given in the output you attached (where it says "Open H2O Flow in browser: http://10.172.243.109:54321 ') into your browser?

cheers,

Lauren

--
You received this message because you are subscribed to the Google Groups "H2O Open Source Scalable Machine Learning - h2ostream" group.
To unsubscribe from this group and stop receiving emails from it, send an email to h2ostream+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Ethan Zou]

Hi Lauren,

I tried http://10.172.243.109:54321 and ec2-52-36-243-54.us-west-2.compute.amazonaws.com:54321, but neither worked.

Ethan

[Michal Malohlava]

Hi Ethan,

there is a problem that Databricks instnaces are living in private network, so you need to expose H2O's instance port 54321 to outside world.
That means you need to somehow make instance which is hosting H2OContext (the same machine which is hosting Spark COntext) visible on public IP.

SO there are two solutions:
 1) One is described in Databricks documentation: assign elastic IP with a private instance which hosts SparkContext. Look for "Assign elastic IP" notebook in your Databricks account

OR

 2) Login into your AWS account, create an EC2 instance and make it colocated with created Databricks ec2 machines (you have to select the same private network), assign a public EIP with it, and run proxy on it which will be connected to a DBC instance which hosts SparkContext.

Please let us know if it works for you!
michal

[Ethan Zou]

Hi Michal,

Thank you for your prompt reply. I tried the first solution but got an error at the very last step. Could you please take a look at my program? I understand it may be a question for Databricks, but they haven't replied to my questions so far. Any suggestion would be much appreciated. 

Followed the "AWS Configuration / Elastic IPAddress" from Databricks, I created an Elastic IP on my own AWS account, provisioned an IAMPolicy and attached to my IAM user. 

gathered ACCESS_KEY, SECRET_KEY, and EIP_ADDRESS. Then I follow the notebook step by step, until the last line. I got an error says Instance not found, but I do see the instance exist. Any suggestions?

Thanks,

Ethan

--
You received this message because you are subscribed to a topic in the Google Groups "H2O Open Source Scalable Machine Learning - h2ostream" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/h2ostream/IqcAb9xZLFg/unsubscribe.
To unsubscribe from this group and all its topics, send an email to h2ostream+...@googlegroups.com.

[rich...@gmail.com]

Hi

Same problem for me I've success to assign EIP to my databricks cluster as described in the guide but H2o context give the same private IP

I've check security group and it's open for the port 54321
Do you have any idea ?