Connect to HDInsight (Apache Hadoop) using SSH
Learn how to use Secure Shell (SSH) to securely connect to Apache Hadoop on Azure HDInsight. For information on connecting through a virtual network, see Azure HDInsight virtual network architecture and Plan a virtual network deployment for Azure HDInsight clusters.
The following table contains the address and port information needed when connecting to HDInsight using an SSH client:
||22||edge node (ML Services on HDInsight)|
||22||edge node (any other cluster type, if an edge node exists)|
<clustername> with the name of your cluster. Replace
<edgenodename> with the name of the edge node.
If your cluster contains an edge node, we recommend that you always connect to the edge node using SSH. The head nodes host services that are critical to the health of Hadoop. The edge node runs only what you put on it. For more information on using edge nodes, see Use edge nodes in HDInsight.
When you first connect to HDInsight, your SSH client may display a warning that the authenticity of the host can't be established. When prompted select 'yes' to add the host to your SSH client's trusted server list.
If you have previously connected to a server with the same name, you may receive a warning that the stored host key does not match the host key of the server. Consult the documentation for your SSH client on how to remove the existing entry for the server name.
Linux, Unix, and macOS systems provide the
scp commands. The
ssh client is commonly used to create a remote command-line session with a Linux or Unix-based system. The
scp client is used to securely copy files between your client and the remote system.
Microsoft Windows doesn't install any SSH clients by default. The
scp clients are available for Windows through the following packages:
OpenSSH Client. This is an optional feature introduced in the Windows 10 Fall Creators Update.
Azure Cloud Shell. The Cloud Shell provides a Bash environment in your browser.
There are also several graphical SSH clients, such as PuTTY and MobaXterm. While these clients can be used to connect to HDInsight, the process of connecting is different than using the
ssh utility. For more information, see the documentation of the graphical client you're using.
Authentication: SSH Keys
SSH keys use public-key cryptography to authenticate SSH sessions. SSH keys are more secure than passwords, and provide an easy way to secure access to your Hadoop cluster.
If your SSH account is secured using a key, the client must provide the matching private key when you connect:
Most clients can be configured to use a default key. For example, the
sshclient looks for a private key at
~/.ssh/id_rsaon Linux and Unix environments.
You can specify the path to a private key. With the
-iparameter is used to specify the path to private key. For example,
ssh -i ~/.ssh/id_rsa firstname.lastname@example.org.
If you have multiple private keys for use with different servers, consider using a utility such as ssh-agent (https://en.wikipedia.org/wiki/Ssh-agent). The
ssh-agentutility can be used to automatically select the key to use when establishing an SSH session.
If you secure your private key with a passphrase, you must enter the passphrase when using the key. Utilities such as
ssh-agent can cache the password for your convenience.
Create an SSH key pair
ssh-keygen command to create public and private key files. The following command generates a 2048-bit RSA key pair that can be used with HDInsight:
ssh-keygen -t rsa -b 2048
You're prompted for information during the key creation process. For example, where the keys are stored or whether to use a passphrase. After the process completes, two files are created; a public key and a private key.
The public key is used to create an HDInsight cluster. The public key has an extension of
The private key is used to authenticate your client to the HDInsight cluster.
You can secure your keys using a passphrase. A passphrase is effectively a password on your private key. Even if someone obtains your private key, they must have the passphrase to use the key.
Create HDInsight using the public key
|Creation method||How to use the public key|
|Azure portal||Uncheck Use cluster login password for SSH, and then select Public Key as the SSH authentication type. Finally, select the public key file or paste the text contents of the file in the SSH public key field.
|Azure PowerShell||Use the
|Azure CLI||Use the
|Resource Manager Template||For an example of using SSH keys with a template, see Deploy HDInsight on Linux with SSH key. The
SSH accounts can be secured using a password. When you connect to HDInsight using SSH, you're prompted to enter the password.
Microsoft does not recommend using password authentication for SSH. Passwords can be guessed and are vulnerable to brute force attacks. Instead, we recommend that you use SSH keys for authentication.
The SSH account password expires 70 days after the HDInsight cluster is created. If your password expires, you can change it using the information in the Manage HDInsight document.
Create HDInsight using a password
|Creation method||How to specify the password|
|Azure portal||By default, the SSH user account has the same password as the cluster login account. To use a different password, uncheck Use cluster login password for SSH, and then enter the password in the SSH password field.
|Azure PowerShell||Use the
|Azure CLI||Use the
|Resource Manager Template||For an example of using a password with a template, see Deploy HDInsight on Linux with SSH password. The
Change the SSH password
For information on changing the SSH user account password, see the Change passwords section of the Manage HDInsight document.
Authentication: Domain-joined HDInsight
If you're using a domain-joined HDInsight cluster, you must use the
kinit command after connecting with SSH local user. This command prompts you for a domain user and password, and authenticates your session with the Azure Active Directory domain associated with the cluster.
You can also enable Kerberos Authentication on each domain joined node (for example, head node, edge node) in order to ssh using the domain account. To do this edit sshd config file:
sudo vi /etc/ssh/sshd_config
uncomment and change
sudo service sshd restart
At any time, in order to verify whether the Kerberos authentication was successful or not, use
For more information, see Configure domain-joined HDInsight.
Connect to nodes
The head nodes and edge node (if there is one) can be accessed over the internet on ports 22 and 23.
When connecting to the head nodes, use port 22 to connect to the primary head node and port 23 to connect to the secondary head node. The fully qualified domain name to use is
clusternameis the name of your cluster.
# Connect to primary head node # port not specified since 22 is the default ssh email@example.com # Connect to secondary head node ssh -p 23 firstname.lastname@example.org
When connecting to the edge node, use port 22. The fully qualified domain name is
edgenodenameis a name you provided when creating the edge node.
clusternameis the name of the cluster.
# Connect to edge node ssh email@example.com
The previous examples assume that you are using password authentication, or that certificate authentication is occurring automatically. If you use an SSH key-pair for authentication, and the certificate is not used automatically, use the
-i parameter to specify the private key. For example,
ssh -i ~/.ssh/mykey firstname.lastname@example.org.
Once connected, the prompt changes to indicate the SSH user name and the node you're connected to. For example, when connected to the primary head node as
sshuser, the prompt is
Connect to worker and Apache Zookeeper nodes
The worker nodes and Zookeeper nodes aren't directly accessible from the internet. They can be accessed from the cluster head nodes or edge nodes. The following are the general steps to connect to other nodes:
Use SSH to connect to a head or edge node:
From the SSH connection to the head or edge node, use the
sshcommand to connect to a worker node in the cluster:
To retrieve a list of the node names, see the Manage HDInsight by using the Apache Ambari REST API document.
If the SSH account is secured using a password, enter the password when connecting.
If the SSH account is secured using SSH keys, make sure that SSH forwarding is enabled on the client.
Another way to directly access all nodes in the cluster is to install HDInsight into an Azure Virtual Network. Then, you can join your remote machine to the same virtual network and directly access all nodes in the cluster.
For more information, see Plan a virtual network for HDInsight.
Configure SSH agent forwarding
The following steps assume a Linux or UNIX-based system, and work with Bash on Windows 10. If these steps do not work for your system, you may need to consult the documentation for your SSH client.
Using a text editor, open
~/.ssh/config. If this file doesn't exist, you can create it by entering
touch ~/.ssh/configat a command line.
Add the following text to the
Host <edgenodename>.<clustername>-ssh.azurehdinsight.net ForwardAgent yes
Replace the Host information with the address of the node you connect to using SSH. The previous example uses the edge node. This entry configures SSH agent forwarding for the specified node.
Test SSH agent forwarding by using the following command from the terminal:
This command returns information similar to the following text:
If nothing is returned, then
ssh-agentisn't running. For more information, see the agent startup scripts information at Using ssh-agent with ssh (http://mah.everybody.org/docs/ssh) or consult your SSH client documentation.
Once you've verified that ssh-agent is running, use the following to add your SSH private key to the agent:
If your private key is stored in a different file, replace
~/.ssh/id_rsawith the path to the file.
Connect to the cluster edge node or head nodes using SSH. Then use the SSH command to connect to a worker or zookeeper node. The connection is established using the forwarded key.
scp utility can be used to copy files to and from individual nodes in the cluster. For example, the following command copies the
test.txt directory from the local system to the primary head node:
scp test.txt email@example.com:
Since no path is specified after the
:, the file is placed in the
sshuser home directory.
The following example copies the
test.txt file from the
sshuser home directory on the primary head node to the local system:
scp firstname.lastname@example.org:test.txt .
scp can only access the file system of individual nodes within the cluster. It cannot be used to access data in the HDFS-compatible storage for the cluster.
scp when you need to upload a resource for use from an SSH session. For example, upload a Python script and then run the script from an SSH session.
For information on directly loading data into the HDFS-compatible storage, see the following documents: