Skip to content

Latest commit

 

History

History
266 lines (173 loc) · 10.6 KB

sql-server-linux-deploy-pacemaker-cluster.md

File metadata and controls

266 lines (173 loc) · 10.6 KB
title description author ms.author ms.reviewer ms.date ms.service ms.subservice ms.topic ms.custom
Deploy a Pacemaker cluster for SQL Server on Linux
Learn to deploy a Linux Pacemaker cluster for a SQL Server Always On availability group (AG) or failover cluster instance (FCI).
rwestMSFT
randolphwest
vanto
07/15/2024
sql
linux
install-set-up-deploy
intro-deployment
linux-related-content

Deploy a Pacemaker cluster for SQL Server on Linux

[!INCLUDE SQL Server - Linux]

This tutorial documents the tasks required to deploy a Linux Pacemaker cluster for a [!INCLUDE ssnoversion-md] Always On availability group (AG) or failover cluster instance (FCI). Unlike the tightly coupled Windows Server / [!INCLUDE ssnoversion-md] stack, Pacemaker cluster creation and availability group (AG) configuration on Linux, can be done before or after installation of [!INCLUDE ssnoversion-md]. The integration and configuration of resources for the Pacemaker portion of an AG or FCI deployment is done after the cluster is configured.

Important

An AG with a cluster type of None does not require a Pacemaker cluster, nor can it be managed by Pacemaker.

[!div class="checklist"]

  • Install the high availability add-on and install Pacemaker.
  • Prepare the nodes for Pacemaker (RHEL and Ubuntu only).
  • Create the Pacemaker cluster.
  • Install the SQL Server HA and SQL Server Agent packages.

Prerequisites

Install SQL Server on Linux.

Install the high availability add-on

Use the following syntax to install the packages that make up the high availability (HA) add-on for each distribution of Linux.

Red Hat Enterprise Linux (RHEL)

  1. Register the server using the following syntax. You're prompted for a valid username and password.

    sudo subscription-manager register

  2. List the available pools for registration.

    sudo subscription-manager list --available

  3. Run the following command to associate RHEL high availability with the subscription

    sudo subscription-manager attach --pool=<PoolID>

    where PoolId is the pool ID for the high availability subscription from the previous step.

  4. Enable the repository to be able to use the high availability add-on.

    sudo subscription-manager repos --enable=rhel-ha-for-rhel-7-server-rpms

  5. Install Pacemaker.

    sudo yum install pacemaker pcs fence-agents-all resource-agents

SUSE Linux Enterprise Server (SLES)

Install the High Availability pattern in YaST or do it as part of the main installation of the server. The installation can be done with an ISO/DVD as a source or by getting it online.

Note

On SLES, the HA add-on gets initialized when the cluster is created.

Ubuntu

sudo apt-get install pacemaker pcs fence-agents resource-agents

Prepare the nodes for Pacemaker (RHEL and Ubuntu only)

Pacemaker itself uses a user created on the distribution named hacluster. The user gets created when the HA add-on is installed on RHEL and Ubuntu.

  1. On each server that will serve as a node of the Pacemaker cluster, create the password for a user to be used by the cluster. The name used in the examples is hacluster, but any name can be used. The name and password must be the same on all nodes participating in the Pacemaker cluster.

    sudo passwd hacluster

  2. On each node that will be part of the Pacemaker cluster, enable and start the pcsd service with the following commands (RHEL and Ubuntu):

    sudo systemctl enable pcsd
sudo systemctl start pcsd

    Then execute the following command to ensure that pcsd is started.

    sudo systemctl status pcsd

  3. Enable the Pacemaker service on each possible node of the Pacemaker cluster.

    sudo systemctl start pacemaker

    On Ubuntu, you see an error:

    pacemaker Default-Start contains no runlevels, aborting.

    This error is a known issue. Despite the error, enabling the Pacemaker service is successful, and this bug will be fixed at some point in the future.

  4. Next, create and start the Pacemaker cluster. There's one difference between RHEL and Ubuntu at this step. While on both distributions, installing pcs configures a default configuration file for the Pacemaker cluster, on RHEL, executing this command destroys any existing configuration and creates a new cluster.

Create the Pacemaker cluster

This section documents how to create and configure the cluster for each distribution of Linux.

Red Hat Enterprise Linux (RHEL)

  1. Authorize the nodes

    sudo pcs cluster auth <Node1 Node2 ... NodeN> -u hacluster

    where NodeX is the name of the node.

  2. Create the cluster

    sudo pcs cluster setup --name <PMClusterName Nodelist> --start --all --enable

    where PMClusterName is the name assigned to the Pacemaker cluster and Nodelist is the list of names of the nodes separated by a space.

SUSE Linux Enterprise Server (SLES)

The process for creating a Pacemaker cluster is different on SLES than it's on RHEL and Ubuntu. The following steps document how to create a cluster with SLES.

  1. Start the cluster configuration process by running the following command on one of the nodes:

    sudo ha-cluster-init

    You might be prompted that NTP isn't configured and that no watchdog device is found. That is fine for getting things up and running. Watchdog is related to fencing a failed node, if you use SLES's built-in fencing that is storage-based. NTP and watchdog can be configured later.

  2. You're prompted to configure Corosync. You're prompted for the network address to bind to, as well as the multicast address and port. The network address is the subnet that you're using; for example, 192.191.190.0. You can accept the defaults at every prompt, or change if necessary.

  3. Next, you're prompted to configure SBD, which is the disk-based fencing. This configuration can be done later if desired. If SBD isn't configured, unlike on RHEL and Ubuntu, stonith-enabled will by default be set to false.

  4. Finally, you're prompted to configure an IP address for administration. This IP address is optional, but functions similar to the IP address for a Windows Server failover cluster (WSFC): it creates an IP address in the cluster to be used for connecting to it via HA Web Konsole (HAWK). This configuration, too, is optional.

  5. Ensure that the cluster is up and running by issuing

    sudo crm status

  6. Change the hacluster password with

    sudo passwd hacluster

  7. If you configured an IP address for administration, you can test it in a browser, which also tests the password change for hacluster.

    :::image type="content" source="media/sql-server-linux-deploy-pacemaker-cluster/image2.png" alt-text="Screenshot of hacluster." lightbox="media/sql-server-linux-deploy-pacemaker-cluster/image2.png":::

  8. On another SLES server that will be a node of the cluster, run

    sudo ha-cluster-join

  9. When prompted, enter the name or IP address of the server that was configured as the first node of the cluster in the previous steps. The server is added as a node to the existing cluster.

  10. Verify the node was added by issuing

    sudo crm status

  11. Change the hacluster password with

    sudo passwd hacluster

  12. Repeat Steps 8-11 for all other servers to be added to the cluster.

Ubuntu

Configuring Ubuntu is similar to RHEL. However, there's one major difference: installing the Pacemaker packages creates a base configuration for the cluster, and enables and starts pcsd. If you try to configure the Pacemaker cluster by following the RHEL instructions exactly, you get an error. To fix this problem, perform the following steps:

  1. Remove the default Pacemaker configuration from each node.

    sudo pcs cluster destroy

  2. Create the cluster

    sudo pcs cluster setup --name <PMClusterName Nodelist> --start --all --enable

    where PMClusterName is the name assigned to the Pacemaker cluster and Nodelist is the list of names of the nodes separated by a space.

Install the SQL Server HA and SQL Server Agent packages

Use the following commands to install the SQL Server HA package and [!INCLUDE ssnoversion-md] Agent, if they aren't installed already. Installing the HA package after installing [!INCLUDE ssnoversion-md] requires a restart of [!INCLUDE ssnoversion-md] for it to be used. These instructions assume that the repositories for the Microsoft packages are already set up, since [!INCLUDE ssnoversion-md] should be installed at this point.

  • If you won't use [!INCLUDE ssnoversion-md] Agent for log shipping or any other use, it doesn't have to be installed, so package mssql-server-agent can be skipped.

  • The other optional packages for [!INCLUDE ssnoversion-md] on Linux, [!INCLUDE ssnoversion-md] Full-Text Search (mssql-server-fts) and [!INCLUDE ssnoversion-md] Integration Services (mssql-server-is), aren't required for high availability, either for an FCI or an AG.

Red Hat Enterprise Linux (RHEL)

sudo yum install mssql-server-ha mssql-server-agent
sudo systemctl restart mssql-server

SUSE Linux Enterprise Server (SLES)

sudo zypper install mssql-server-ha mssql-server-agent
sudo systemctl restart mssql-server

Ubuntu

sudo apt-get install mssql-server-ha mssql-server-agent
sudo systemctl restart mssql-server

Next step

In this tutorial, you learned how to deploy a Pacemaker cluster for SQL Server on Linux. You learned how to:

[!div class="checklist"]

  • Install the high availability add-on and install Pacemaker.
  • Prepare the nodes for Pacemaker (RHEL and Ubuntu only).
  • Create the Pacemaker cluster.
  • Install the SQL Server HA and SQL Server Agent packages.

To create and configure an availability group for SQL Server on Linux, see:

[!div class="nextstepaction"] Create and configure an availability group for SQL Server on Linux.