YugabyteDB Anywhere FAQ

Contents

YugabyteDB Anywhere
Installation
Data nodes

Node agent

YugabyteDB Anywhere

What is YugabyteDB Anywhere?

YugabyteDB Anywhere (previously known as Yugabyte Platform and YugaWare) is a private database-as-a-service, used to create and manage YugabyteDB universes and clusters. YugabyteDB Anywhere can be used to deploy YugabyteDB in any public or private cloud.

You deploy and manage your YugabyteDB universes using the YugabyteDB Anywhere UI.

See also YugabyteDB Anywhere at yugabyte.com.

How do I report a security vulnerability?

Follow the steps in the vulnerability disclosure policy to report a vulnerability to our security team. The policy outlines our commitments to you when you disclose a potential vulnerability, the reporting process, and how Yugabyte will respond.

Installation

How does YugabyteDB Anywhere installation work?

YugabyteDB Anywhere first needs to be installed on a machine. The next step is to configure YugabyteDB Anywhere to work with public and/or private clouds. In the case of public clouds, YugabyteDB Anywhere spawns the machines to orchestrate bringing up the data platform. In the case of private clouds, you add the nodes you want to be a part of the data platform into YugabyteDB Anywhere.

You install YugabyteDB Anywhere using a standalone installer that you download from Yugabyte.

Replicated end of life

YugabyteDB Anywhere was previously installed using Replicated. However, YugabyteDB Anywhere will end support for Replicated installation at the end of 2024. You can migrate existing Replicated YugabyteDB Anywhere installations using YBA Installer. See Migrate from Replicated.

YugabyteDB Anywhere distributes and installs YugabyteDB on the hosts identified to run the data nodes. Because the YugabyteDB software is already packaged into existing artifacts, the data node does not require any Internet connectivity.

For instructions on installing YugabyteDB Anywhere, refer to Install YugabyteDB Anywhere.

What are the OS requirements and permissions to run YugabyteDB Anywhere?

For a list of operating systems supported by YugabyteDB Anywhere, see Operating system support. YugabyteDB Anywhere doesn't support ARM architectures (but can be used to deploy universes to ARM-based nodes).

For Replicated-based installations, YugabyteDB Anywhere supports operating systems supported by Replicated. Replicated only supports Linux-based operating systems. The Linux OS should be 3.10+ kernel, 64-bit, and ready to run docker-engine 1.7.1 - 17.06.2-ce (with 17.06.2-ce being the recommended version). For a complete list of operating systems supported by Replicated, see Supported Operating Systems.

YugabyteDB Anywhere also requires the following:

  • Connectivity to the Internet, either directly or via an HTTP proxy.
  • The following ports open on the platform host: 443 (HTTPS access to the YugabyteDB Anywhere UI), 22 (SSH).
  • Attached disk storage (such as persistent EBS volumes on AWS): 100 GB SSD minimum.
  • A YugabyteDB Anywhere license file from Yugabyte.
  • Ability to connect from the YugabyteDB Anywhere host to all YugabyteDB data nodes via SSH.

For a complete list of prerequisites, refer to YBA prerequisites.

Data nodes

What are the requirements to run YugabyteDB data nodes?

Prerequisites for YugabyteDB data nodes are listed in the YugabyteDB Deployment checklist.

How does the YugabyteDB Anywhere UI interact with YugabyteDB data nodes?

The YugabyteDB Anywhere UI creates a passwordless SSH connection to interact with the data nodes.

Can I access the database machines that get spawned in public clouds?

Yes, you have access to all machines spawned. The machines are spawned by YugabyteDB Anywhere. YugabyteDB Anywhere runs on your machine in your region/data center. If you have configured YugabyteDB Anywhere to work with any public cloud (such as AWS or GCP), it will spawn YugabyteDB nodes using your credentials on your behalf. These machines run in your account, but are created and managed by YugabyteDB Anywhere on your behalf. You can log on to these machines any time. The YugabyteDB Anywhere UI additionally displays metrics per node and per universe.

How many machines do I need to try out YugabyteDB Anywhere against my load?

You need the following:

  • One server to install YugabyteDB Anywhere on.
  • A minimum number of servers for the data nodes as determined by the replication factor (RF). For example, one server for RF=1, and 3 servers in case of RF=3.
  • A server to run the load tests on.

Typically, you can saturate a database server (or three in case of RF=3) with just one large enough test machine running a synthetic load tester that has a light usage pattern. YugabyteDB ships with some synthetic load-testers, which can simulate a few different workloads. For example, one load tester simulates a time series or IoT-style workload and another does a stock-ticker like workload. But if you have a load tester that emulates your planned usage pattern, you can use that.

Can I control the properties (such as VPC, IOPS, tenancy, and so on) of the machines YugabyteDB Anywhere spins up?

Yes, you can control what YugabyteDB Anywhere is spinning up. For example:

  • You can choose if YugabyteDB Anywhere should spawn a new VPC with peering to the VPC on which application servers are running (to isolate the database machines into a separate VPC) AWS, or ask it to reuse an existing VPC.

  • You can choose dedicated IOPs EBS drives on AWS and specify the number of dedicated IOPS you need.

YugabyteDB Anywhere also allows creating these machines out of band and importing them as an on-premises install.

Node agent

What is a node agent?

Node agent is an RPC service running on a YugabyteDB node, and is used to manage communication between YugabyteDB Anywhere and the nodes in universes. It includes the following functions:

  • Invoke shell commands directly on the remote host, similar to running commands over SSH. Like SSH, the agent also does shell-login such that any previous command modifying the environment in the resource files (for example, ~/.bashrc) gets reflected in the subsequent command.
  • Invoke procedures or methods on the node agent.
  • Additionally, for on-premises (manually provisioned nodes) deployments, node agent also functions as a utility to run preflight checks, and add node instances.

How is node agent installed on a YugabyteDB node?

Installation depends on how the nodes are provisioned.

Currently, there are three ways to provision a node, depending on the level of access provided to YBA, as follows:

  • Automatic provisioning, where an SSH user with sudo access for the node is provided to YBA (for example, the ec2-user for an AWS EC2 instance).

  • Assisted manual provisioning via a script (provision_instance.py), where YBA doesn't have access to an SSH user with sudo access, but you can run the script interactively in YBA, providing parameters for credentials for the SSH user with sudo access.

  • Fully manual provisioning, where neither you nor YBA has access to an SSH user with sudo access. In this case, only a local (non-SSH) user with sudo access is available, and you must follow a series of manual steps to provision the node.

For cloud (AWS, GCP, and Azure) and automatic on-premises (sudo access provided) providers, agents are automatically installed on each universe node during provisioning using SSH. After the node agent is installed, all the legacy SSH calls are replaced with node agent calls. Just like an SSH daemon, node agent is run as a root user to perform provisioning of the nodes. After node provisioning, you can revoke the SSH access, but it's recommended to retain access for debugging.

For manually provisioned on-premises providers, node agent is installed on each node either using a script or manually as part of the manual provisioning process.

Why does YBA prompt for SSH details if the node agent is installed manually and can replace SSH?

When creating an on-premises provider, you are prompted to provide SSH credentials, which are used during provisioning of automatically provisioned providers. After provisioning and adding the instances to the provider (including installing the node agent), YugabyteDB Anywhere no longer requires SSH or sudo access to nodes.

If you are manually provisioning nodes, these credentials aren't needed to provision nodes.

However, SSH keys are still required to connect to the node for debugging purposes (by navigating to universe > Nodes > Actions > Connect for example).

If you don't want to provide YBA with an SSH key for a manually provisioned on-premises provider (because you can log in to the nodes over SSH for debugging outside of YBA), then you can provide a dummy SSH key in the provider configuration.

What is registration and unregistration of node agent?

Node agent is a secure service that authenticates every remote call from YBA. Registration is the process of:

  • making YBA aware of the node agent service; and
  • establishing trust between the node agent and YBA.

No provider-level details are needed for registration.

Unregistration is the process of removing the node agent entry from YBA such that there is no further communication. In effect, unregistration makes YBA and the node forget each other.

Why does node agent installation ask for provider and other details during on-prem manual node agent setup?

Node agent is used to run pre-flight checks on the node during various day-0 and day-2 operations. These checks need information like the non-root user's home directory, expected port number for Node exporter, the NTP servers, and so on. These are attributes configured with a provider. As a result, to run these pre-flight checks the node agent needs to be configured to a provider in YBA, and these details are needed to make the node agent aware of the YBA provider which the node will become a part of.

How do I move a node provisioned for one provider to a different provider?

In v2.18.6 and later, moving a node from one provider to another does not require unregistering the node agent, as node agents aren't linked to providers.

To change the provider of a node, follow the procedure in Reconfigure a node agent.

As long as the IP does not change, the node agent does not try to register again.

Note that first removing the node instance from the provider is very important, because after the provider configuration information changes in the configuration file as part of running the config command, the node agent will no longer be able to find the node to delete it, as the scope is always with the current provider and availability zone.

How does a node agent perform preflight checks?

Prior to adding a node that you have provisioned as an instance to your provider, you run a preflight check to determine if the node satisfies the requirements for YugabyteDB.

A node agent does the following when the preflight-check command is run:

  1. Runs the preflight_check script.
  2. Collects the output and converts it to a well-formatted JSON payload.
  3. Sends the payload to YBA for validation. YBA has preflight check threshold values defined in the runtime configuration of the provider. The prefix path for the configuration key name is yb.node_agent.preflight_checks. The values can be changed if needed.

How do I disable a node agent?

You can disable node agents of a provider any time by setting the yb.node_agent.client.enabled Provider Configuration for the provider to false.

This disables all node agents for the provider, which falls back to SSH, using credentials provided during provider creation.

How do I change the node agent default port of 9070?

Use the yb.node_agent.server.port Provider Configuration.

The change is reflected only on newly created node agents.

How does YBA determine that a node instance record maps to a node agent entry if they are not related?

YBA uses the IP address to identify a node instance. The IP can be a DNS from YBA version 2.18.5 and later.

For YBA versions 2.18.6 or 2.20.2, a bind address that defaults to the IP will be added in case a DNS is supplied and the node agent has to listen to a specific interface IP.

How does YBA clean up node agents?

For providers where node agents are installed automatically (that is, all providers except on-premises with manual provisioning), YBA removes the node agent entry when it releases the node back to the node instances pool of the provider. You can choose to leave the node agent service running, but when the node is again re-used, the node agent is re-installed.