Using load balancing for highly available applications

This tutorial explains how to use load balancing with a regional managed instance group to redirect traffic away from busy or unavailable VM instances, allowing you to provide high availability even during a zonal outage.

A regional managed instance group distributes an application on multiple instances across multiple zones. A global load balancer directs traffic across multiple regions via a single IP address. By using both of these services to distribute your application across multiple zones, you can help ensure that your application is available even in extreme cases, like a zonal disruption.

Load balancers can be used to direct a variety of traffic types. This tutorial shows you how to create a global load balancer that directs external HTTP traffic, but much of the content of this tutorial is still relevant to other types of load balancers. To learn about other types of traffic that can be directed with a load balancer, see Types of Cloud Load Balancing.

This tutorial includes detailed steps for launching a web application on a regional managed instance group, configuring network access, creating a load balancer for directing traffic to the web application, and observing the load balancer by simulating a zonal outage. Depending on your experience with these features, this tutorial takes about 45 minutes to complete.

Objectives

• Launch a demo web application on a regional managed instance group.
• Configure a global load balancer that directs HTTP traffic across multiple zones.
• Observe the effects of the load balancer by simulating a zonal outage.

Costs

This tutorial uses billable components of Google Cloud including:

• Compute Engine

Use the pricing calculator to generate a cost estimate based on your projected usage. New Google Cloud users might be eligible for a free trial.

Before you begin

1. Sign in to your Google Account.

   If you don't already have one, sign up for a new account.

2. In the Cloud Console, on the project selector page, select or create a Cloud project.

   Note: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.

   Go to the project selector page

3. Make sure that billing is enabled for your Google Cloud project. Learn how to confirm billing is enabled for your project.

Application architecture

The application includes the following Compute Engine components:

• VPC network: a virtual network within Google Cloud that can provide global connectivity via its own routes and firewall rules.
• Firewall rule: a Google Cloud firewall lets you allow or deny traffic to your instances.
• Instance template: a template used to create each VM instance in the managed instance group.
• Regional managed instance group: a group of VM instances running the same application across multiple zones.
• Global static external IP address: a static IP address that is accessible on external networks and can be attached to a global resource.
• Global load balancer: a load balancer that allows backend instances to be distributed across multiple regions. Use a global load balancer when your users need access to the same applications and content, and you want to provide access using a single anycast IP address.
• Health check: a policy used by the load balancer to evaluate the responsiveness of the application on each VM instance.

Launching the web application

This tutorial uses a web application that is stored on GitHub. If you would like learn more about how the application was implemented, see the Google Cloud GitHub repository.

Launch the web application on every VM in an instance group by including a startup script in an instance template. Additionally, run the instance group in a dedicated VPC network to keep this tutorial's firewall rules from interfering with any existing resources running in your project.

Create a VPC network

Using a VPC network protects existing resources in your project from being affected by the resources that you will create for this tutorial. A VPC network is also required to restrict incoming traffic so that it must go through the load balancer.

Create a VPC network to encapsulate the firewall rules for the demo web application:

1. Go to the VPC networks page in the Cloud Console.
   Go to the VPC networks page
2. Click Create VPC Network.
3. Under Name, enter web-app-vpc.

4. Set Subnet creation mode to Automatic.

   Note: Use auto mode network for fast, convenient set up. In production, it's recommended that you use custom mode networks instead. For more information, see Considerations for automode networks.

5. At the bottom of the page, click Create.

Wait until the VPC network has finished being created before continuing.

Create a firewall rule

After the VPC network is created, set up a firewall rule to allow HTTP traffic to the VPC network:

Note: For convenience, create a wide-access firewall rule. But, for production applications, you should implement firewall rules that follow the security best practice of least privilege. For more information, see Networking and security best practices.

1. Go to the Firewalls page in the Cloud Console.
   Go to the Firewalls page
2. Click Create firewall rule.
3. Under Name, enter allow-web-app-http.
4. Set Network to web-app-vpc.
5. Under Targets, select All instances in the network.
6. Set Source filter to IP ranges.
7. Under Source IP ranges, enter 0.0.0.0/0 to allow access for all IP addresses.
8. Under Ports and protocols, select Specified protocols and ports.
   Check tcp and enter 80 to allow access for HTTP traffic.
9. Click Create.

Create an instance template

Create a template that you will use to create a group of VM instances. Each instance created from the template launches a demo web application via a startup script.

1. Go to the Instance templates page in the Cloud Console.
   Go to the Instance templates page
2. Click Create instance template.
3. Under Name, enter load-balancing-web-app-template.
4. Under Machine configuration, set the Machine type to f1-micro.
5. Under Boot disk, set the Image to Debian GNU/Linux 9 (stretch).
6. Under Firewall, select Allow HTTP traffic.
7. Click Management, security, disks, networking, sole tenancy to reveal the advanced settings.

8. Click the Management tab. Under Automation, enter the following Startup script.

   sudo apt update && sudo apt -y install git gunicorn3 python3-pip
   git clone https://github.com/GoogleCloudPlatform/python-docs-samples.git
   cd python-docs-samples/compute/managed-instances/demo
   sudo pip3 install -r requirements.txt
   sudo gunicorn3 --bind 0.0.0.0:80 app:app --daemon
   

   The script gets, installs, and launches the web application at instance startup.

9. Click the Networking tab. Under Network, select web-app-vpc. This forces each instance created with this template to run on the previously created network.

10. Click Create.

Wait until the template has finished being created before continuing.

Create a regional managed instance group

To run the web application, use the instance template to create a regional managed instance group:

1. Go to the Instance groups page in the Cloud Console.
   Go to the Instance groups page
2. Click Create instance group.
3. Under Name, enter load-balancing-web-app-group.

4. Under Location, select Multiple zones.

   Pro Tip: To ensure your application is available during extreme events, like zonal outages, Compute Engine recommends that you distribute your application across multiple zones.

5. Under Region, select us-central1.

6. Click the Configure zones drop-down menu to reveal Zones. Select the following zones:

   • us-central1-b
   • us-central1-c
   • us-central1-f

7. Under Instance template, select load-balancing-web-app-template.

8. Under Autoscaling, select Off.

9. Set Number of instances to 6.

   Pro Tip: When creating a regional managed instance group, Compute Engine recommends that you provision enough instances so that, if all of the instances in any one zone are unavailable, the remaining instances still meet the minimum number of instances that you require. However, provisioning more instances than you need might incur additional costs. For more information, see Selecting instance group size to ensure availability.

10. Under Instance redistribution, select On.

11. Under Autohealing and Health check, select No health check.

12. Click Create. This redirects you back to the Instance groups page.

    Note: You might need to wait a few minutes until all of the instances in the group are running and ready to display the web application.

13. To verify that your instances are running the demo web application correctly:

    1. From the Instance groups page, click load-balancing-web-app-group to see the instances in that group.

    2. Under External IP, click an IP address to connect that instance. A new browser tab opens displaying the demo web application:

       

       Note: If you are unable to connect to the web application after a few minutes, check the instance status and network settings:

       • Verify that the instance group is ready. After the instance is running, you might need to wait a few minutes for the startup script to finish running.
       • Verify that the instance group is running on the web-app-vpc network and that the instance template has Allow HTTP traffic enabled. Then, verify that the allow-web-app-http firewall rule was created correctly.

       When you are done, close the browser tab for the demo web application.

Configuring the load balancer

To use a load balancer to direct traffic to your web application, you must reserve an external IP address to receive all incoming traffic. Then, create a load balancer that accepts traffic from that IP address and redirects that traffic to the instance group.

Reserve a static IP address

Use a global static external IP address to provide the load balancer with a single point of entry for receiving all user traffic. Compute Engine preserves static IP addresses even if you change or delete any affiliated Google Cloud resources. This allows the web application to always have the same entry point, even if other parts of the web application might change.

1. Go to the External IP addresses page in the Cloud Console.
   Go to the External IP addresses page
2. Click Reserve static address.
3. Under Name, enter web-app-ipv4.
4. Set IP version to IPv4.
5. Set Type to Global.
6. Click Reserve.

Create a load balancer

This section explains the steps required to create a global load balancer that directs HTTP traffic.

This load balancer uses a frontend to receive incoming traffic and a backend to distribute this traffic to healthy instances. Because the load balancer is made of multiple components, this task is divided into several parts:

• Backend configuration
• Frontend configuration
• Review and finalize

Complete all the steps to create the load balancer.

Note: For simplicity, this tutorial uses an HTTP load balancer. To learn how to support HTTPS and HTTP/2, see Creating content-based HTTP(S) load balancing. For other types of traffic, see Choosing a load balancer.

1. Go to the Create a load balancer page in the Cloud Console.
   Go to the Create a load balancer page
2. Under HTTP(S) Load Balancing, click Start configuration.
3. Under Internet facing or internal only, select From Internet to my VMs. Then, click Continue.
4. For the Name of the load balancer, enter web-app-load-balancer.

Backend configuration

1. In the left panel of the New HTTP(S) load balancer page, click Backend configuration.
2. Click Create or select backend services & backend buckets to open a drop-down menu. Click Backend services, then click Create a backend service.
3. In the new window, for the Name of the backend application, enter web-app-backend.
4. Set Instance group to load-balancing-web-app-group.
5. Set Port numbers to 80. This allows HTTP traffic between the load balancer and the instance group.
6. Under Balancing mode, select Utilization.
7. Click Done to create the backend.

8. Create the health check for the backend of the load balancer:

   Pro Tip: Health checks are used for both load balancing and autohealing, but for different purposes:

   • Health checks for load balancing are used for detecting unresponsive instances and directing traffic away from them.
   • Health checks for autohealing are used for detecting and recreating failed instances.

   Use separate health checks for load balancing and for autohealing. Using the same health check for these services would remove the distinction between unresponsive instances and failed instances, causing unnecessary latency and/or unavailability for your users. For more information, see Health check concepts.

   1. Under Health check, select Create a health check (or Create another health check) from the drop-down menu. A new window opens.
   2. In the new window under Name, enter web-app-load-balancer-check.
   3. Set the Protocol to HTTP.
   4. Under Port, enter 80.
   5. For this tutorial, set the Request path to /health, which is a path that the demo web application is set up to respond to.

   6. Set the following Health criteria:

      1. Set Check interval to 3 seconds. This defines the amount of time from the start of one probe to the start of the next one.
      2. Set Timeout to 3 seconds. This defines the amount of time that Google Cloud waits for a response to a probe. Its value must be less than or equal to the check interval.
      3. Set Healthy Threshold to 2 consecutive successes. This defines the number of sequential probes that must succeed in order for the instance to be considered healthy.
      4. Set Unhealthy Threshold to 2consecutive failures. This defines the number of sequential probes that must fail in order for the instance to be considered unhealthy.

      Pro Tip: For information about refining the Check interval and Timeout values for your own application, see How health checks work. To refine the Healthy Threshold and Unhealthy Threshold values for your own application, use the following criteria:

      • If traffic is being directed to instances with high latency even when faster instances are available, decrease the threshold for traffic by making your health check more aggressive.
      • If traffic is being directed away from instances with low latency, increase the threshold for traffic by making your health check more conservative.

      For detailed information about optimizing and measuring latency, see Optimizing Application Latency with Load Balancing

   7. Click Save and continue to create the health check.

9. Click Create to create the backend service.

Frontend configuration

1. In the left panel of the New HTTP(S) load balancer page, click Frontend configuration.
2. On the Frontend configuration page, under Name, enter web-app-ipv4-frontend.
3. Set the Protocol to HTTP.
4. Set the IP version to IPv4.
5. Set the IP address to web-app-ipv4.
6. Set the Port to 80.
7. Click Done to create the frontend.

Review and finalize

1. Verify your load balancing settings before creating the load balancer:

   1. In the left panel of the New HTTP(S) load balancer page, click Review and finalize.

   2. On the Review and finalize page, verify the following Backend settings:

      • The Backend service is web-app-backend.
      • The Endpoint protocol is HTTP.
      • The Health check is web-app-load-balancer-check.
      • The Instance group is load-balancing-web-app-group.

   3. On the same page, verify that Frontend uses an IP address with a Protocol of HTTP.

2. In the left panel of the New HTTP(S) load balancer page, click Create to finish creating the load balancer.

You might need to wait a few minutes for the load balancer to finish being created.

Simulating a zonal outage

You can observe the functionality of the load balancer by simulating the widespread unavailability of a zonal outage. This simulation works by forcing all of the instances located in a specified zone to report an unhealthy status on the /health request path. When these instances report an unhealthy status, they fail the load balancing health check, prompting the load balancer to stop directing traffic to these instances.

1. Monitor which zones the load balancer is directing traffic to.

   1. Open a terminal by using Cloud Shell from the Cloud Console.

      Open Cloud Shell

      Cloud Shell opens on the bottom of the Cloud Console. It can take a few seconds for the session to initialize.

      Pro Tip:

      You can open the Cloud Shell from any Cloud Console page using the Activate Cloud Shell button, which is in the top right corner of every Cloud Console page.

   2. Save the static external IP address of your load balancer:

      1. Get the external IP address from the frontend forwarding rule of the load balancer by entering the following command in your terminal:

         gcloud compute forwarding-rules list | grep web-app-ipv4-frontend
         

         Copy the EXTERNAl_IP_ADDRESS from the output:

         web-server-ipv4-frontend    EXTERNAl_IP_ADDRESS    TCP    web-app-load-balancer-target-proxy
         

      2. Create a local bash variable:

         export LOAD_BALANCER_IP=EXTERNAl_IP_ADDRESS
         

         where EXTERNAl_IP_ADDRESS is the external IP address that you copied.

   3. To monitor which zones the load balancer is directing traffic to, run the following bash script:

      while true
      do
              BODY=$(curl -s "$LOAD_BALANCER_IP")
              NAME=$(echo -n "$BODY" | grep "load-balancing-web-app-group" | perl -pe 's/.+?load-balancing-web-app-group-(.+?)<.+/\1/')
              ZONE=$(echo -n "$BODY" | grep "us-" | perl -pe 's/.+?(us-.+?)<.+/\1/')
      
              echo $ZONE
      done
      

      This script continuously attempts to connect to the web application via the IP address for the frontend of the load balancer, and outputs which zone the web application is running from for each connection.

      The resulting output should include zones us-central1-b, us-central1-c, and us-central1-f:

      us-central1-f
      us-central1-b
      us-central1-c
      us-central1-f
      us-central1-f
      us-central1-c
      us-central1-f
      us-central1-c
      us-central1-c
      

      Keep this terminal open.

      Note: This monitor should run continuously. But, you can stop it at any time by pressing Control+C in the terminal.

2. While your monitor is running, begin simulating the zonal outage.

   1. In Cloud Shell, open a second terminal session by clicking the Add button.

   2. Create a local bash variable for the project ID:

      export PROJECT_ID=PROJECT_ID
      

      where PROJECT_ID is the project ID for your current project, which is displayed on each new line in the Cloud Shell:

      user@cloudshell:~ (PROJECT_ID)$
      

   3. Create a local bash variable for the zone that you want to disable. To simulate a failure of zone us-central1-f, use the following command:

      export DISABLE_ZONE=us-central1-f
      

      Then, run the following bash script. This script causes the demo web application instances in the disabled zone to output unhealthy responses to the load balancer health check. Unhealthy responses prompt the load balancer to direct traffic away from these instances.

      export MACHINES=$(gcloud --project=$PROJECT_ID compute instances list --filter="zone:($DISABLE_ZONE)" --format="csv(name,networkInterfaces[0].accessConfigs[0].natIP)" | grep "load-balancing-web-app-group")
      for i in $MACHINES;
      do
        NAME=$(echo "$i" | cut -f1 -d,)
        IP=$(echo "$i" | cut -f2 -d,)
        echo "Simulating zonal failure for zone $DISABLE_ZONE, instance $NAME"
        curl -q -s "http://$IP/makeUnhealthy" >/dev/null --retry 2
      done
      

      After a short delay, the load balancer stops directing traffic to the unhealthy zones, so the output from the first terminal window stops listing zone us-central1-f:

      us-central1-c
      us-central1-c
      us-central1-c
      us-central1-b
      us-central1-b
      us-central1-c
      us-central1-b
      us-central1-c
      us-central1-c
      

      This indicates that the load balancer is directing traffic only to the healthy, responsive instances.

      Note: Optionally, you can repeat this step to simulate failures of zones us-central1-b and us-central1-c.

      Keep both terminals open.

   4. In the second terminal, create a local bash variable for the zone that you want to restore. To restore traffic to zone us-central1-f, use the following command:

      export ENABLE_ZONE=us-central1-f
      

      Then, run the following bash script. This script causes the demo web application instances in the enabled zone to output healthy responses to the load balancer health check. Healthy responses prompt the load balancer to begin distributing traffic back toward these instances.

      export MACHINES=$(gcloud --project=$PROJECT_ID compute instances list --filter="zone:($ENABLE_ZONE)" --format="csv(name,networkInterfaces[0].accessConfigs[0].natIP)" | grep "load-balancing-web-app-group")
      for i in $MACHINES;
      do
        NAME=$(echo "$i" | cut -f1 -d,)
        IP=$(echo "$i" | cut -f2 -d,)
        echo "Simulating zonal restoration for zone $ENABLE_ZONE, instance $NAME"
        curl -q -s "http://$IP/makeHealthy" >/dev/null --retry 2
      done
      

      After a few minutes, the output from the first terminal window gradually lists zone us-central1-fagain:

      us-central1-b
      us-central1-b
      us-central1-c
      us-central1-f
      us-central1-c
      us-central1-c
      us-central1-b
      us-central1-c
      us-central1-f
      

      This indicates that the load balancer is directing incoming traffic to all zones again.

      Note: If you also disabled zone us-central1-b or zone us-central1-c, you can repeat this step to restore traffic to them.

      Close both terminals when you have finished.

(Optional) Restricting incoming traffic

When you created the regional managed instance group, you could access each instance directly through its external ephemeral IP address. However, now that you have prepared a load balancer and static external IP address, you might want to modify the network firewall so that incoming traffic must go through the load balancer.

If you want to restrict incoming traffic to the load balancer, modify the network firewall to disable the ephemeral external IP address for each instance.

Note: The zonal outage simulation will stop working after you restrict incoming traffic to the load balancer. To make all the instances in a specific zone healthy or unhealthy, the simulation needs to specify which instances it connects to.

1. Edit the firewall rule to restrict HTTP traffic so that the web application can only be accessed via the load balancer:

   1. Go to the Firewalls page in the Cloud Console.
      Go to the Firewalls page
   2. Under Name, click allow-web-app-http.
   3. Click Edit.

   4. Modify the Source IP ranges to only allow health check probes:

      1. Delete 0.0.0.0/0.
      2. On the same line, enter 130.211.0.0/22 and press Tab.
      3. On the same line, enter 35.191.0.0/16and press Tab.
      Note: Health check probes from the load balancer come from addresses in the ranges 130.211.0.0/22 and 35.191.0.0/16. For this tutorial, your health check uses the HTTP protocol, so the firewall rule should allow connections to port 80. For more information on firewall rules for health checks, see firewall rules for load balancing.

   5. Click Save.

2. Verify that you cannot connect to the web application via the ephemeral external IP address for a specific instance:

   1. Go to the Instance groups page in the Cloud Console.
      Go to the Instance groups page

   2. Click load-balancing-web-app-group to see the instances in that group.

   3. Under External IP, click an IP address to connect that instance. A new browser tab opens, but the web application does not open. (Eventually, the page will show a timeout error).

      When you are done, close the browser tab for the instance.

3. Verify that you can connect to the web application via the load balancer:

   1. Go to the Load balancing page in the Cloud Console.
      Go to the Load balancing page
   2. Under Name, click web-app-load-balancer to expand the load balancer you just created.

   3. To connect to the web-app via the external static IP addresses, look under Frontend and IP:Port, and copy the IP address. Then, open a new browser tab and paste the IP address into the address bar. This should display the demo web application:

      

      Notice that, whenever you refresh the page, the load balancer connects to different instances in different zones. This happens because you are not connecting to an instance directly; you are connecting to the load balancer, which selects the instance you are redirected to.

      When you are done, close the browser tab for the demo web application.

Cleaning up

After you've finished the load balancing tutorial, you can clean up the resources that you created on Google Cloud so they won't take up quota and you won't be billed for them in the future. The following sections describe how to delete or turn off these resources.

If you created a separate project for this tutorial, delete the entire project. Otherwise, if the project has resources that you want to keep, only delete the resources created in this tutorial.

Deleting the project

Caution: Deleting a project has the following effects:

• Everything in the project is deleted. If you used an existing project for this tutorial, when you delete it, you also delete any other work you've done in the project.
• Custom project IDs are lost. When you created this project, you might have created a custom project ID that you want to use in the future. To preserve the URLs that use the project ID, such as an appspot.com URL, delete selected resources inside the project instead of deleting the whole project.

If you plan to explore multiple tutorials and quickstarts, reusing projects can help you avoid exceeding project quota limits.

1. In the Cloud Console, go to the Manage resources page.

   Go to the Manage resources page

2. In the project list, select the project that you want to delete and then click Delete .
3. In the dialog, type the project ID and then click Shut down to delete the project.

Deleting specific resources

Deleting the load balancer

1. Go to the Load balancing page in the Cloud Console.
   Go to the Load balancing page
2. Click the checkbox next to web-app-load-balancer.
3. Click Delete at the top of the page.
4. In the new window, select all checkboxes. Then, click Delete load balancer and selected resources to confirm the deletion.

Deleting the static external IP address

Note: Wait until the load balancer is deleted before deleting the static external IP address.

1. Go to the External IP addresses page in the Cloud Console.
   Go to the External IP addresses page
2. Click the checkbox next to web-app-ipv4.
3. Click Release static address at the top of the page. In the new window, click Delete to confirm the deletion.

Deleting the instance group

Note: Wait until the load balancer is deleted before deleting the instance group.

1. In the Cloud Console, go to the Instance groups page.

   Go to the Instance groups page

2. Click the checkbox for your load-balancing-web-app-group instance group.
3. Click Delete to delete the instance group.

Deleting the instance template

Note: You must finish deleting the instance group before deleting the instance template. You cannot delete an instance template if a managed instance group is using it.

1. Go to the Instance Templates page in the Cloud Console.

   Go to the Instance templates page

2. Click the checkbox next to load-balancing-web-app-template.

3. Click Delete at the top of the page. In the new window, click Delete to confirm the deletion.

Deleting the VPC network

Note: You must finish deleting the instance group before deleting the VPC network. You cannot delete a VPC network if other resources still uses it.

1. Go to the VPC networks page in the Cloud Console.

   Go to the VPC networks page

2. Click web-app-vpc.

3. Click Delete at the top of the page. In the new window, click Delete to confirm the deletion.

What's next

• Try another tutorial:
  • Using autohealing for highly available applications.
  • Using autoscaling for highly scalable applications.
• Learn more about Managed Instance Groups.
• Learn more about Load Balancing.
• Learn more about Optimizing Application Latency with Load Balancing.
• Learn more about Designing Robust Systems.
• Learn more about Building Scalable and Resilient Web Applications on Google Cloud.