Running H2O on AWS cluster
Prerequisites:
- AWS_ACCESS_KEY_ID
- AWS_SECRET_ACCESS_KEY
- Key pair .pem
- Network security profile
Click on my Security credentials
Click Access keys then create access keys
Create key pair, by going to ec2 dashboard then Network and security then create key pairs. Give it a name and save the key file .pem
Create a security group from ec2 dashboard then network&security then security groups
Create security group, save its name and add rule…All traffic
Working procedure:
First change the permissions rights of the key using
chmod 600 /path/to/key/key_name.pem
On the linux machine, open ‘h2o-cluster-launch-instances.py’ and change the following
- Access key ID, access key, and path to the key file created
os.environ['AWS_ACCESS_KEY_ID'] = 'AKIAJHHVRLCYUDSB2EQQ'
os.environ['AWS_SECRET_ACCESS_KEY'] = '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'
os.environ['AWS_SSH_PRIVATE_KEY_FILE'] = '/home/mohamed/Desktop/aws/created_key.pem'
- Name of the key and security group name
keyName = 'created_key'
securityGroupName = 'SecurityDisabled' - Name and number of the instances
numInstancesToLaunch = 2
instanceType = 't2.micro'
instanceNameRoot = 'h2o-instance'
- Might need to check if the amiId is still the same as in the console, make sure it is the same region
- Open ‘h2o-cluster-test-ssh.sh’, ‘h2o-cluster-distribute-flatfile.sh’, ‘h2o-cluster-download-h2o.sh’, ‘h2o-cluster-start-h2o.sh’ and add the path to the key file
AWS_SSH_PRIVATE_KEY_FILE="/home/mohamed/Desktop/aws/created_key.pem" - Open ‘h2o-cluster-distribute-aws-credentials.sh’ and add credentials and path to key
AWS_ACCESS_KEY_ID="AKIAJHHVRLCYUDSB2EQQ"
AWS_SECRET_ACCESS_KEY="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"
AWS_SSH_PRIVATE_KEY_FILE="/home/mohamed/Desktop/aws/created_key.pem"
Now all files are ready. Either run the command
./run-all.sh
Or run one by one for debugging
./h2o-cluster-launch-instances.py
./h2o-cluster-download-h2o.sh
./h2o-cluster-distribute-aws-credentials.sh
./h2o-cluster-start-h2o.sh
Now open the browser on any of the public IPs of the clusters on port 54321
111.222.333.444:54321
Open Admin→Cluster status and make sure the number of nodes are there
No comments:
Post a Comment