Prepare nodes for on-premises deployment

For on-premises deployments of Yugabyte universes, you need to import nodes that can be managed by Yugabyte Platform. This page outlines the steps required to prepare these YugabyteDB nodes for on-premises deployments.

Ports

The following ports must be opened for intra-cluster communication (they do not need to be exposed to your application, only to other nodes in the cluster and the platform node):

• 7100 - Master RPC
• 9100 - TServer RPC

The following ports must be exposed for intra-cluster communication, and you should additionally expose these ports to administrators or users monitoring the system, as these ports provide valuable diagnostic troubleshooting and metrics:

• 9300 - Prometheus metrics
• 7000 - Master HTTP endpoint
• 9000 - TServer HTTP endpoint
• 11000 - YEDIS API
• 12000 - YCQL API
• 13000 - YSQL API

The following nodes must be available to your application or any user attempting to connect to the YugabyteDB, in addition to intra-node communication:

• 5433 - YSQL server
• 9042 - YCQL server
• 6379 - YEDIS server

For more information on ports used by YugabyteDB, refer to Default ports.

Preparing nodes

To prepare nodes for on premises deployment:

1. Ensure that the YugabyteDB nodes conform to the requirements outlined in the deployment checklist. This checklist also gives an idea of recommended instance types across public clouds.

2. Install the prerequisites and verify the system resource limits as described in system configuration.

3. Ensure you have ssh access to the machine and root access (or the ability to run sudo; the sudo user can require a password but having passwordless access is desirable for simplicity and ease of use).

4. Verify that you can ssh into this node (from your local machine if the node has a public address).

   $ ssh -i your_private_key.pem ssh_user@node_ip
   

The following actions are performed with sudo access:

• Create the yugabyte:yugabyte user + group.

• Set the home directory to /home/yugabyte.

• Create the prometheus:prometheus user + group.

  Tip

  If you're using the LDAP directory for managing system users, you can pre-provision Yugabyte and Prometheus users:

  • The yugabyte user should belong to the yugabyte group.

  • Set the home directory for the yugabyte user (default /home/yugabyte) and ensure the directory is owned by yugabyte:yugabyte. The home directory is used during cloud provider configuration.

  • The Prometheus username and the group can be user-defined. You enter the custom user during cloud provider configuration.

• Ensure you can schedule Cron jobs with Crontab. Cron jobs are used for health monitoring, log file rotation, and cleanup of system core files.

  Tip

  For any 3rd party cron scheduling tools, you can disable Crontab and add these cron entries:

  # Ansible: cleanup core files hourly
  0 * * * * /home/yugabyte/bin/clean_cores.sh
  # Ansible: cleanup yb log files hourly
  5 * * * * /home/yugabyte/bin/zip_purge_yb_logs.sh
  # Ansible: Check liveness of master
  */1 * * * * /home/yugabyte/bin/yb-server-ctl.sh master cron-check || /home/yugabyte/bin/yb-server-ctl.sh master start
  # Ansible: Check liveness of tserver
  */1 * * * * /home/yugabyte/bin/yb-server-ctl.sh tserver cron-check || /home/yugabyte/bin/yb-server-ctl.sh tserver start
  

  Disabling Crontab creates alerts after the universe is created, but they can be ignored. But you need to ensure cron jobs are set appropriately for the platform to work as expected.

• Verify that Python 2.7 is installed.

• Enable core dumps and set ulimits.

  *       hard        core        unlimited
  *       soft        core        unlimited
  

• Configure SSH as follows:

  • Disable sshguard.
  • Set UseDNS no in /etc/ssh/sshd_config (disables reverse lookup, which is used for auth; DNS is still useable).

• Set vm.swappiness to 0.

• Set mount path permissions to 0755.