Mar 092016
 

In this step by step guide I will take you through all steps required to configure a highly available, 2-node MySQL cluster (plus witness server) in Microsoft Azure IaaS (Infrastructure as a Service).  The guide includes both screenshots, shell commands and code snippets as appropriate.  I assume that you are somewhat familiar with Microsoft Azure and already have an Azure account with an associated subscription.  If not, you can sign up for a free account today.  I’m also going to assume that you have basic linux system administration skills as well as understand basic failover clustering concepts like Virtual IPs, etc.

Disclaimer: Azure is a rapidly moving target.  It’s getting better and better every day!  As such, features/screens/buttons are bound to change over time so your experience may vary slightly from what you’ll see below.  While this guide will show you how to make a MySQL database highly available, you could certainly adapt this information and process to protect other applications or databases, like SAP, Oracle, PostgreSQL, NFS file servers, and more.

These are the high level steps to create a highly available MySQL database within Microsoft Azure IaaS:

  1. Create a Resource Group
  2. Create a Virtual Network
  3. Create a Storage Account
  4. Create Virtual Machines in an Availability Set
  5. Set VM Static IP Addresses
  6. Add a Data Disk to cluster nodes
  7. Create Inbound Security Rule to allow VNC access
  8. Linux OS Configuration
  9. Install and Configure MySQL
  10. Install and Configure Cluster
  11. Create an Internal Load Balancer
  12. Test Cluster Connectivity

Overview

This article will describe how to create a cluster within a single Azure region.  The cluster nodes (node1, node2 and the witness server) will reside in an Availability Set (3 different Fault Domains and Update Domains), thanks to the new Azure Resource Manager (ARM). We will be creating all resources using the new Azure Resource Manager.

The configuration will look like this:

Cluster-Diagram

The following IP addresses will be used:

  • node1: 10.0.0.4
  • node2: 10.0.0.5
  • witness: 10.0.0.6
  • virtual/”floating” IP: 10.0.0.99
  • MySQL port: 3306

Create a Resource Group

First, create a Resource Group.  Your resource group will end up containing all of the various objects related to our cluster deployment: virtual machines, virtual network, storage account, etc.  Here we will call our newly created Resource Group “cluster-resources”.


resource group1

Be mindful when selecting your region.  All of your resources will need to reside within the same region.  Here, we’ll be deploying everything into the “West US” region:

resource group2

Create a Virtual Network (VNet)

Next, create a Virtual Network.  A Virtual Network is an isolated network within the Azure cloud that is dedicated to you.  You have full control over things like IP address blocks and subnets, routing, security policies (i.e. firewalls), DNS settings, and more.  You will be launching your Azure Iaas virtual machines (VMs) into your Virtual Network.

virtual network1

Make sure you select Resource Manager as the deployment model anytime you are given the option:

virtual network2

Give your new Virtual Network a name (“virtual-network”) and make sure you select the resource group that was created in the previous step (“cluster-resources”).  Your Virtual Network needs to reside in the same region as your Resource Group.  We will leave the IP Address and Subnet values as default.

virtual network3

Create a Storage Account

Before you provision any Virtual Machines, you’ll need to create a Storage Account where they will be stored.

storage account1

Again, make sure you select Resource Manager as the deployment model anytime you are given the option:

storage account2

Next, give your new storage account a name.  The storage account name must be unique across *ALL* of Azure.  (Every object that you store in Azure Storage has a unique URL address. The storage account name forms the subdomain of that address.)  In this example I call my storage account “linuxclusterstorage” but you’ll need to select something different as you setup your own.

Select a storage Type based on your requirements and budget.  For the purposes of this guide, I selected “Standard-LRS” (i.e. Locally Redundant) to minimize cost.

Make sure your new Storage Account is added to the Resource Group you created in Step 1 (“cluster-resources”)  in the same Location (“West US” in this example):

storage account3

Create Virtual Machines in an Availability Set

We will be provisioning 3 Virtual Machines in this guide.  The first two VMs (I’ll call them “node1” and “node2”) will function as cluster nodes with the ability to bring the MySQL database and it’s associated resources online.  The 3rd VM will act as the cluster’s witness server for added protection against split-brain.

To ensure maximum availability, all 3 VMs will be added to the same Availability Set, ensuring that they will end up in different Fault Domains and Update Domains.

Create “node1” VM

Create your first VM (“node1”).  In this guide we will be using CentOS 6.X:

create vm1

Make sure you use the Resource Manager deployment model.  It should be selected by default:

create vm2

Give the VM a hostname (“node1”) and username/password that will later be used to SSH into the system.  Make sure you add this VM to your Resource Group (“cluster-resources”) and that it resides in the same region as all of your other resources:

create vm3

Next, choose your instance size.  For more information on the various instance sizes available, click here.

For the purposes of this guide, I’m using “A3 Standard” for Node1 and Node2, to minimize cost since this won’t be running a production workload.  I used an even smaller “A1 Standard” size for the witness server.  Select the instance size that makes most sense for you.

create vm4

If you want to be able to connect into the VM from the outside world, set a Public IP address.  I did this so I can later SSH and VNC into the system

create vm5

IMPORTANT: By default, your VM won’t be added to an Availability Set.  On the Settings screen during make sure you create a new Availability Set, we’ll call “cluster-availability-set”.  Azure Resource Manager (ARM) allows your to create Availability Sets with 3 Fault Domains.  The default values here are fine:

create vm6

Review your VM properties and click OK to create your first VM:

create vm7

Create “node2” and “witness” VMs

Repeat the steps above twice to create two more VMs.  I created another “A3 Standard” size VM called “node2” and an “A1 Standard” size VM called “witness”.

The only difference here is that you’ll be ADDING these VMs to the Availability Set (“cluster-availability-set”) we just created:

create vm8

It may take a little while for your 3 VMs to provision.  Once complete, you’ll see your VMs listed on the Virtual Machines screen within your Azure Portal:

create vm9

Set VM Static IP Addresses

The VMs will be set with the following IP addresses:

  • node1: 10.0.0.4
  • node2: 10.0.0.5
  • witness: 10.0.0.6

Repeat this step for each VM.  Select your VM and edit the Network Interfaces

 

static ip address1

Select the network interface associated with the VM, and edit IP addresses.  Select “Static” and specify the desired IP address:

static ip address2

Add a Data Disk to cluster nodes

Next, we will need to add a extra disk to of our cluster nodes (“node1” and “node2”).  This disk will store our MySQL databases and the later be replicated between nodes.

Note: You do NOT need to add an extra disk to the “witness” node.  Only “node1” and “node2”.

Edit your VM, select Disks and then attach a new disk:

second disk1

Select a disk type (Standard or Premium SSD)  and size based on your workload.  Here I create a 10GB Standard disk on both of my cluster nodes.  As far as Host caching goes, “None” or “Read only” caching is fine.  I do not recommend using “Read/Write” as there is potential for data loss:

second disk2

Create Inbound Security Rule to allow VNC access

If your VM is part of a Network Security Group (NSG), which by default it likely is unless you disabled it during VM creation, the only port open in the “Azure firewall” is SSH (port 22).  Later in the guide, I’ll be using VNC to access the desktop of “node1” and configure the cluster using a GUI.  Create an Inbound Security Rule to open up VNC access.  In this guide port 5902 is used.  Adjust this according based on your VNC configuration.

Virtual Machines -> (select node1) -> Network interfaces -> (select NIC) -> Network security group -> (select the NSG) -> Inbound security rules -> Add


inbound security rule1

Linux OS Configuration

Here is where we will leave the Azure Portal for a little while and get our hands dirty on the command line, which as a Linux administrator you should be used to by now.  You aren’t given the root password to your Linux VMs in Azure, so once you login as the user specified during VM creation, use the “sudo” command to gain root privileges:

$sudo su -

Edit /etc/hosts

Unless you have already have a DNS server setup, you’ll want to create host file entries on all 3 servers so that they can properly resolve each other by name

Add the following lines to the end of your /etc/hosts file:

10.0.0.4    node1
10.0.0.5    node2
10.0.0.6    witness
10.0.0.99   mysql-vip

Disable SELinux

Edit /etc/sysconfig/linux and set “SELINUX=disabled”:

# vi /etc/sysconfig/selinux

# This file controls the state of SELinux on the system.
# SELINUX= can take one of these three values:
#     enforcing - SELinux security policy is enforced.
#     permissive - SELinux prints warnings instead of enforcing.
#     disabled - No SELinux policy is loaded.
SELINUX=disabled
# SELINUXTYPE= can take one of these two values:
#     targeted - Targeted processes are protected,
#     mls - Multi Level Security protection.
SELINUXTYPE=targeted

Configure iptables so that cluster the Virtual IP will work

IMPORTANT: In order to get connectivity to the cluster Virtual IP to work, and also monitoring of the IP resource, a few iptables rules need to be setup.  Note: 10.0.0.99 is the Virtual IP we’ll be using in our cluster, and 3306 is the default port used my MySQL.

On node1 (10.0.0.4), run the following commands:

# iptables --flush
# iptables -t nat -A PREROUTING -p tcp --dport 3306 -j DNAT --to-destination 10.0.0.99:3306
# iptables -t nat -A POSTROUTING -p icmp -s 10.0.0.99 -j SNAT --to-source 10.0.0.4
# service iptables save
# chkconfig iptables on

On Node2 (10.0.0.5), run the following commands:

# iptables --flush
# iptables -t nat -A PREROUTING -p tcp --dport 3306 -j DNAT --to-destination 10.0.0.99:3306
# iptables -t nat -A POSTROUTING -p icmp -s 10.0.0.99 -j SNAT --to-source 10.0.0.5
# service iptables save
# chkconfig iptables on

Install and Configure VNC (and related packages)

In order to access the GUI of our linux servers, to later configure our cluster, install VNC server on your cluster node.  In my setup I only did this on “node1”

# yum install tigervnc-server xterm
# vncpasswd
# vi /etc/sysconfig/vncservers

      VNCSERVERS="2:root"
      VNCSERVERARGS[2]="-geometry 1024x768"

# service vncserver start
# chkconfig vncserver on

Test connectivity by opening a VNC client on your laptop/desktop, and connecting to the Public IP of your cluster node

Reboot Cluster Nodes

Reboot your cluster nodes so that SELinux is disabled, and the 2nd disk you previously added is detected. Only “node1” and “node2” need to be rebooted.

Partition and Format the “data” disk

In Step 6 of this guide (“Add a Data Disk to cluster nodes”) we did just that….added an extra disk to each cluster node to store the application data we will be protecting.  In this case it happens to be MySQL databases.

In Azure IaaS, Linux Virtual Machines use the following arrangement for disks:

  • /dev/sda – OS disk
  • /dev/sdb – temporary disk
  • /dev/sdc – 1st data disk
  • /dev/sdd – 2nd data disk
  • /dev/sdj – 8th data disk

The disk we added in Step 6 of this guide should appear as /dev/sdc.  You can run the “fdisk -l” command to verify.  You’ll see that /dev/sda (OS) and /dev/sdb (temporary) already have disk partitions and are being used.

# fdisk -l

Disk /dev/sdb: 306.0 GB, 306016419840 bytes
255 heads, 63 sectors/track, 37204 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0xd3920649

Device Boot      Start         End      Blocks   Id  System
/dev/sdb1   *           1       37205   298842112   83  Linux

Disk /dev/sdc: 10.7 GB, 10737418240 bytes
255 heads, 63 sectors/track, 1305 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x00000000

Disk /dev/sda: 32.2 GB, 32212254720 bytes
255 heads, 63 sectors/track, 3916 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x000c23d3

Device Boot      Start         End      Blocks   Id  System
/dev/sda1   *           1        3789    30432256   83  Linux
/dev/sda2            3789        3917     1024000   82  Linux swap / Solaris

 

Here I will create a partition (/dev/sdc1), format it, and mount it at the default location for MySQL, which is /var/lib/mysql.  Perform the following steps on BOTH “node1” and “node2”:

# fdisk /dev/sdc
Command (m for help): n
Command action
e   extended
p   primary partition (1-4)
p
Partition number (1-4): 1
First cylinder (1-1305, default 1): <enter>
Using default value 1
Last cylinder, +cylinders or +size{K,M,G} (1-1305, default 1305): <enter>
Using default value 1305
 
Command (m for help): w
The partition table has been altered!
Calling ioctl() to re-read partition table.
Syncing disks.
[root@node1 ~]#

# mkfs.ext4 /dev/sdc1
# mkdir /var/lib/mysql

On node1, mount the filesystem:

# mount /dev/sdc1 /var/lib/mysql

Install and Configure MySQL

Next, install install the MySQL packages, initialize a sample database, and set “root” password for MySQL.

On “node1”:

# yum -y install mysql mysql-server
# /usr/bin/mysql_install_db --datadir="/var/lib/mysql/" --user=mysql
# mysqld_safe --user=root --socket=/var/lib/mysql/mysql.sock --port=3306 --datadir=/var/lib/mysql --log &
#
# # NOTE: This next command allows remote connections from ANY host.  NOT a good idea for production!
# echo “update user set Host='%' where Host='node1'; flush privileges | mysql mysql
#
# #Set MySQL's root password to 'SIOS'
# echo "update user set Password=PASSWORD('SIOS') where User='root'; flush privileges" | mysql mysql

Create a MySQL configuration file. We will place this on the data disk  (that will later be replicated – /var/lib/mysql/my.cnf).  Example:

# vi /var/lib/mysql/my.cnf

[mysqld]
datadir=/var/lib/mysql
socket=/var/lib/mysql/mysql.sock
pid-file=/var/lib/mysql/mysqld.pid
user=root
port=3306
# Disabling symbolic-links is recommended to prevent assorted security risks
symbolic-links=0
 
[mysqld_safe]
log-error=/var/log/mysqld.log
pid-file=/var/run/mysqld/mysqld.pid
 
[client]
user=root
password=SIOS

Delete the original MySQL configuration file, located in /etc, if it exists:

# rm /etc/my.cnf

On “node2”:

On “node2”, you ONLY need to install the MySQL packages.  The other steps aren’t required:

[root@node2 ~]# yum -y install mysql mysql-server

Install and Configure the Cluster

At this point, we are ready to install and configure our cluster.  SIOS Protection Suite for Linux (aka SPS-Linux) will be used in this guide as the clustering technology.  It provides both high availability failover clustering features (LifeKeeper) as well as real-time, block level data replication (DataKeeper) in a single, integrated solution.  SPS-Linux enables you to deploy a “SANLess” cluster, aka a “shared nothing” cluster meaning that cluster nodes don’t have any shared storage, as is the case with Azure VMs.

Install SIOS Protection Suite for Linux

Perform the following steps on ALL 3 VMs (node1, node2, witness):

Download the SPS-Linux installation image file (sps.img) and and obtain either a trial license or purchase permanent licenses.  Contact SIOS for more information.

You will loopback mount it and run the “setup” script inside, as root (or first “sudo su -” to obtain a root shell)

For example:

# mkdir /tmp/install
# mount -o loop sps.img /tmp/install
# cd /tmp/install
# ./setup

During the installation script, you’ll be prompted to answer a number of questions.  You will hit Enter on almost every screen to accept the default values.  Note the following exceptions:

  • On the screen titled “High Availability NFS” you may select “n” as we will not be creating a highly available NFS server
  • Towards the end of the setup script, you can choose to install a trial license key now, or later. We will install the license key later, so you can safely select “n” at this point
  • In the final screen of the “setup” select the ARKs (Application Recovery Kits, i.e. “cluster agents”) you wish to install from the list displayed on the screen.
    • The ARKs are ONLY required on “node1” and “node2”.  You do not need to install on “witness”
    • Navigate the list with the up/down arrows, and press SPACEBAR to select the following:
      • lkDR – DataKeeper for Linux
      • lkSQL – LifeKeeper MySQL RDBMS Recovery Kit
    • This will result in the following additional RPMs installed on “node1” and “node2”:
      • steeleye-lkDR-9.0.2-6513.noarch.rpm
      • steeleye-lkSQL-9.0.2-6513.noarch.rpm

Install Witness/Quorum package

The Quorum/Witness Server Support Package for LifeKeeper (steeleye-lkQWK) combined with the existing failover process of the LifeKeeper core allows system failover to occur with a greater degree of confidence in situations where total network failure could be common. This effectively means that failovers can be done while greatly reducing the risk of “split-brain” situations.

Install the Witness/Quorum rpm on all 3 nodes (node1, node2, witness):

# cd /tmp/install/quorumrpm -Uvh steeleye-lkQWK-9.0.2-6513.noarch.rpm

On ALL 3 nodes (node1, node2, witness), edit /etc/default/LifeKeeper, set

NOBCASTPING=1

On ONLY the Witness server (“witness”), edit /etc/default/LifeKeeper, set

WITNESS_MODE=off/none

Install a License key

On all 3 nodes, use the “lkkeyins” command to install the license file that you obtained from SIOS:

# /opt/LifeKeeper/bin/lkkeyins <path_to_file>/<filename>.lic

Start LifeKeeper

On all 3 nodes, use the “lkstart” command to start the cluster software:

# /opt/LifeKeeper/bin/lkstart

Set User Permissions for LifeKeeper GUI

On all 3 nodes, edit /etc/group and add the “tony” user (or whatever username you specified during VM creation) to the “lkadmin” group to grant access to the LifeKeeper GUI.  By default only “root” is a member of the group, and we don’t have the root password in :

# vi /etc/group

lkadmin:x:1001:root,tony

Open the LifeKeeper GUI

Make a VNC connection to the Public IP address of node1.  Based on the VNC and Inbound Security Rule configuration from above, you would connect to <Public_IP>:2 using the VNC password you specified earlier.  Once logged in, open a terminal window and run the LifeKeeper GUI using the following command:

# /opt/LifeKeeper/bin/lkGUIapp &

You will be prompted to connect to your first cluster node (“node1”).  Enter the linux userid and password specified during VM creation:

lk-gui-connect1

Next, connect to both “node2” and “witness” by clicking the “Connect to Server” button highlighted in the following screenshot:

lk-gui-connect2

You should now see all 3 servers in the GUI, with a green checkmark icon indicating they are online and healthy:

lk-gui-connect3

Create Communication Paths

Right-click on “node1” and select Create Comm Path

comm path1

Select BOTH “node2” and “witness” and then follow the wizard.  This will create comm paths between:

  • node1 & node2
  • node1 & witness

comm path2

A comm path still needs to be created between node2 & witness.   Right click on “node2” and select Create Comm Path.  Follow the wizard and select “witness” as the remote server:

comm path3

At this point the following comm paths have been created:

  • node1 <—> node2
  • node1 <—> witness
  • node2 <—> witness

The icons in front of the servers have changed from a green “checkmark” to a yellow “hazard sign”.  This is because we only have a single communication path between nodes.

If the VMs had multiple NICs (information on creating Azure VMs with multiple NICs can be found here, but won’t be covered in this article), you would create redundant comm paths between each server.

comm path4

To remove the warning icons, go to the View menu and de-select “Comm Path Redundancy Warning”:

comm path5

Result:

comm path6

Verify Communication Paths

Use the “lcdstatus” command to view the state of cluster resources.  Run the following commands to verify that you have correctly created comm paths on each node to the other two servers involved:

# /opt/LifeKeeper/bin/lcdstatus -q -d node1

MACHINE  NETWORK ADDRESSES/DEVICE   STATE     PRIO

node2    TCP     10.0.0.4/10.0.0.5  ALIVE        1

witness  TCP     10.0.0.4/10.0.0.6  ALIVE        1

#/opt/LifeKeeper/bin/lcdstatus -q -d node2

MACHINE  NETWORK ADDRESSES/DEVICE   STATE     PRIO

node1    TCP     10.0.0.5/10.0.0.4  ALIVE        1

witness  TCP     10.0.0.5/10.0.0.6  ALIVE        1

#/opt/LifeKeeper/bin/lcdstatus -q -d witness

 

MACHINE  NETWORK ADDRESSES/DEVICE   STATE     PRIO

node1    TCP     10.0.0.6/10.0.0.4  ALIVE        1

node2    TCP     10.0.0.6/10.0.0.5  ALIVE        1

Create a Data Replication cluster resource (i.e. Mirror)

Next, create a Data Replication resource to replicate the /var/lib/mysql partition from node1 (source) to node2 (target).  Click the “green plus” icon to create a new resource:

data replication1

Follow the wizard with these selections:

Please Select Recovery Kit:  Data Replication
Switchback Type: intelligent
Server: node1
Hierarchy Type: Replicate Exiting Filesystem
Existing Mount Point: /var/lib/mysql
Data Replication Resource Tag: datarep-mysql
File System Resource Tab: /var/lib/mysql
Bitmap File: (default value)
Enable Asynchronous Replication:  No

After the resource has been created, the “Extend” (i.e. define backup server) wizard will appear.  Use the following selections:

Target Server: node2
Switchback Type: Intelligent
Template Priority: 1
Target Priority: 10
Target Disk: /dev/sdc1
Data Replication Resource Tag: datarep-mysql
Bitmap File: (default value)
Replication Path: 10.0.0.4/10.0.0.5
Mount Point: /var/lib/mysql
Root Tag: /var/lib/mysql

The cluster will look like this:

data replication2

Create Virtual IP

Next, create a Virtual IP cluster resource.  Click the “green plus” icon to create a new resource:

virtual ip1

Follow the wizard with to create the IP resource with these selections:

Select Recovery Kit: IP
Switchback Type: Intelligent
IP Resource: 10.0.0.99
Netmask: 255.255.255.0
Network Interface: eth0
IP Resource Tag: ip-10.0.0.99

Extend the IP resource with these selections:

Switchback Type: Intelligent
Template Priority: 1
Target Priority: 10
IP Resource: 10.0.0.99
Netmask: 255.255.255.0
Network Interface: eth0
IP Resource Tag: ip-10.0.0.99

Configure a Ping List for the IP resource

By default, SPS-Linux monitors the health of IP resources by performing a broadcast ping.  In many virtual and cloud environments, broadcast pings don’t work.  In a previous step, we set “NOBCASTPING=1” in /etc/default/LifeKeeper to turn off broadcast ping checks. Instead, we will define a ping list.  This is a list of IP addresses to be pinged during IP health checks for this IP resource.   In this guide, we will add the witness server (10.0.0.6) to our ping list.

Right click on the IP resource (ip-10.0.0.99) and select Properties:

ping list1

You will see that initially, no ping list is configured for our 10.0.0.0 subnet.   Click “Modify Ping List”:

ping list2

Enter “10.0.0.6” (the IP address of our witness server), click “Add address” and finally click “Save List”:

ping list3

You will be returned to the IP properties panel, and can verify that 10.0.0.6 has been added to the ping list.  Click OK to close the window:

ping list4

Create the MySQL resource hierarchy

Next, create a MySQL cluster resource.  The MySQL resource is responsible for stopping/starting/monitoring of your MySQL database.  To create, click the “green plus” icon to create a new resource:

Follow the wizard with to create the IP resource with these selections:

Select Recovery Kit: MySQL Database
Switchback Type: Intelligent
Server: node1
Location of my.cnf: /var/lib/mysql
Location of MySQL executables: /usr/bin
Database Tag: mysql

Extend the IP resource with the following selections:

Target Server: node2
Switchback Type: intelligent
Template Priority: 1
Target Priority: 10

As a result, your cluster will look as follows.  Notice that the Data Replication resource was automatically moved underneath the database (dependency automatically created) to ensure it’s always brought online before the database:

mysql-resource1

Create a Dependency between the IP resource and the MySQL Database resource

Create a dependency between the IP resource and the MySQL Database resource so that they failover together as a group.  Right click on the “mysql” resource and select “Create Dependency”:

create-dependency1

On the following screen, select the “ip-10.0.0.99” resource as the dependency.  Click Next and continue through the wizard:

create-dependency2

At this point the SPS-Linux cluster configuration is complete.  The resource hierarchy will look as follows:

create-dependency3

 

Create an Internal Load Balancer

If this was a typical on-premises cluster using either physical or virtual servers, you’d be done at this point.  Clients and Applications would connect into the Virtual IP of the cluster (10.0.0.99) to reach the active node.  In Azure, this doesn’t work without some additional configuration.

You will notice that you can’t connect to the Virtual IP from any server other than the node that is currently active.  Most cloud providers, including Azure, do not allow or support gratuitous ARPs which is the reason you can’t connect to the Virtual IP directly.

To workaround this, Azure provides a feature were you can setup an Internal Load Balancer (ILB).  Essentially, when you connect to the IP address of the ILB (which we will actually set to be the same as the cluster’s Virtual IP – 10.0.0.99) you are routed to the currently active cluster node.

Create a Load Balancer:

internal load balancer1

Give it a name, select “Internal” as the scheme, make sure your virtual network and subnet are properly selected, and assign a static IP that is the same as the cluster’s Virtual IP address.  In this example it’s 10.0.0.99:

internal load balancer2

Next, add a backend pool behind the load balancer.  This how you place the two cluster VMs behind this load balancer

internal load balancer3

Select both of your cluster nodes (node1, node2) and add them to the Backend Pool:

internal load balancer4

Once saved, expand the backend pool (called “ILBBackEnd” here) and you’ll see both VMs underneath along with their status and IPs.  It may take a few seconds before the screen updates:

internal load balancer5

Next, configure a probe for your ILB.  The probe checks the health of a service behind the ILB to determine which node to route traffic to.  Here we will specify port 3306, which is the default for MySQL:

internal load balancer6

Finally, complete the ILB configuration by creating a Load Balancing Rule.   TCP, Port 3306, and make sure you select “Enabled” for “Floating IP (direct server return)”:

internal load balancer7

Test Cluster Connectivity

At this point, all of our Azure and Cluster configurations are complete!

Cluster resources are currently active on node1:

create-dependency3

SSH into the witness server, “sudo su -” to gain root access.   Install the mysql client if needed:

[root@witness ~]# yum -y install mysql

Test MySQL connectivity to the cluster:

[root@witness ~]# mysql --host=10.0.0.99 mysql -u root -p

Execute the following MySQL query to display the hostname of the active cluster node:

mysql> select @@hostname;
+------------+
| @@hostname |
+------------+
| node1      |
+------------+
1 row in set (0.00 sec)
mysql>

Using LifeKeeper GUI, failover from Node1 -> Node2″.  Right click on the mysql resource underneath node2, and select “In Service…”:

failover1

After failover:

failover2

After failover has completed, re-run the MySQL query.  You’ll notice that the MySQL client has detected that the session was lost (during failover) and automatically reconnects:

Execute the following MySQL query to display the hostname of the active cluster node, verifying that now “node2” is active:

mysql> select @@hostname;
ERROR 2006 (HY000): MySQL server has gone away
No connection. Trying to reconnect...
Connection id:    48
Current database: mysql
+------------+
| @@hostname |
+------------+
| node2      |
+------------+
1 row in set (0.56 sec)
mysql>

 

  3 Responses to “Step-By-Step: How to configure a Linux failover cluster in Microsoft Azure IaaS without shared storage #azure #sanless”

  1. […] Disclaimer: Azure is a rapidly moving target.  And here I’m working with a public preview version of SQL Server for Linux.  As such, features/screens/buttons are bound to change before SQL v.Next is officially released so your experience may vary slightly from what you’ll see below.  While this guide will show you how to make a SQL Server for Linux database highly available, you could certainly adapt this information and process to protect other applications or databases, like I have written about before here (MySQL example). […]

  2. Simply wanna remark that you have a very nice website, I enjoy the layout it really stands out.

  3. always i used to read smaller articles which as well clear their motive, and
    that is also happening with this piece of writing which I am reading at this place.

 Leave a Reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

(required)

(required)