Docs
How does horizontal Autoscaling work
This article explains how Horizontal autoscaling works, what the requirements are and how you can enable it.
How does it work
With horizontal autoscaling, additional Hypernodes will be added automatically to your main Hypernode at peak times. The extra Hypernodes will be used as PHP fpm workers. Since there’s no configuration or IP changes needed, Horizontal autoscaling has zero-downtime.
For Horizontal autoscaling you can configure some conditions. If the conditions are met from the configured settings in the Control Panel, Horizontal autoscaling will be triggered. You can configure the following conditions:
CPU Load Threshold: CPU load that is continuously monitored by the autoscaling agent. The percentage value you see is the actual load divided by the number of CPUs you have. For example, if your actual load is 2, but you have 4 CPUs, your CPU load percentage is 50%. The default setting is 70%.
Minimum Duration in Minutes: Minimum amount of time the server is overloaded before autoscaling gets triggered. We monitor your CPU load every minute; autoscaling will be triggered once the CPU load condition is met for the minimum amount of time set in the Control Panel. The default setting is 15 minutes.
Horizontal autoscaling process
Upscaling
Once the configured conditions are met and autoscaling is triggered, we spin up extra Hypernode(s) which meets the extra calculated amount of resources you need. We create a snapshot of your main Hypernode and attach that snapshot to the extra provisioned Hypernodes. All the services, such as MySQL, Elasticsearch and Redis, keep running on your original Hypernode. We add the additional Hypernodes as PHP-fpm workers and will handle the incoming requests. The load of the different autoscaling nodes will be balanced from your original Hypernode.
If the Hypernodes remains overloaded post-autoscaling, the extra needed resources will be added to handle the traffic. The autoscaling trigger is determined based on these criteria. The greater of either:
User-defined minimum duration for CPU threshold surpassing: Allows time to assess newly allocated resources’ impact on CPU performance.
A cooldown period of 15 minutes: Ensures a minimum interval for reassessment when the user-defined duration is less than 15 minutes.
This approach selects the longer duration between user-defined settings or the fallback duration before initiating the next autoscaling event. It ensures adequate time for evaluating resource changes on CPU performance. Autoscaling is capped at the biggest plan available on each cloud provider.
Downscaling
If the monitor agent notices that the used resources are below the configured thresholds, the downscale operation will begin. This will lead to removing the additional servers that are no longer needed.
Downscale will happen by detaching the autoscaling instances from the main Hypernode one by one. After a downscale operation has been completed, there will be a cooldown of 15 minutes to monitor the behaviour of the newly available resources before downscaling again.
If the amount of used resources is still below the configured threshold, another downscale operation will be triggered.
If there is need for extra resources instead, extra resources will be added again.
Horizontal autoscaling requirements
Horizontal autoscaling is available on all the Falcon cloud plans (OpenStack).
Next to the provider, horizontal autoscaling does have a couple of additional requirements.
We will divide them between Hypernode-specific and Application-specific requirements
Hypernode Specific Requirements
Operating system
Horizontal autoscaling is only available on Hypernodes with the Debian Bookworm operating system.
If you would like to upgrade the operating system of your Hypernode, feel free to contact our support team for help.
Make sure the Hypernode is a production plan
For now, we don’t support horizontal autoscaling for development plans.
Enable and configure Varnish
Varnish is a requirement, because it’s essential to deliver cached pages from the load balancer instead of the application servers when scaled up. The benefits are:
Pages in the cache are served faster, because it is one less network hop
Saves network bandwidth
Saves CPU usage for both load balancer and application servers
Page cache is not affected by scaling up or down
You can check if Varnish is enabled on your Hypernode by running
app@abcdef-example-magweb-cmbl:~$ hypernode-systemctl settings varnish_enabled
varnish_enabled is set to value True
If Varnish is not enabled, you can enable Varnish by following the documentation If Varnish is enabled on your Hypernode, your Magento store should also be configured to make use of varnish.
Verify that your VCL has the health probe disabled. While we do support the health probe, we do not recommend to use it in a basic setup.
The health probe can be removed by removing the .probe from your backend default section in the VCL file.
If you do, however, want to use the health probe, make sure that you set the backend host to 127.0.0.1 instead of localhost.
Additionally, make sure the IP range 10.0.0.0/24 is allowed under the acl purge section in the Varnish VCL file. The acl purge section should look like the following:
acl purge {
"localhost";
"10.0.0.0/24";
}
Make sure to use MySQL 5.7 or higher
MySQL 5.7 or higher is a requirement, because those versions support auto negotiation of TLS. TLS is an essential part for horizontal autoscaling, because the application servers will connect to the MySQL instance over the internet instead of wireguard.
You can check the enabled MySQL version by running the following command:
app@abcdef-example-magweb-cmbl:~$ hypernode-systemctl settings mysql_version
mysql_version is set to value 8.0
If your MySQL version is still set to 5.6, you can consider upgrading the MySQL version to a supported version for autoscaling.
After the version validation, please verify the MySQL host is set to mysqlmaster. You can verify this by running cat app/etc/env.php | grep -i mysql | grep -i host from the magento root.
You should see something similar to 'host' => 'mysqlmaster',. If this is not the case please make sure the database connection host is set to mysqlmaster instead of localhost or 127.0.0.1 in the magento configuration file at <magento_root>/app/etc/env.php.
Disable supervisor services
Make sure supervisor is disabled and that there are no supervisor services configured.
For now, we don’t support having supervisor services, because we can’t guarantee everything works when scaling up. If you do think your use case should be supported, please let us know!
Disable Podman services
Make sure podman is disabled and that there are no podman services running.
For now, we don’t support having podman containers/services, because we can’t guarantee everything works when scaling up. If you do think your use case should be supported, please let us know!
Configure hostnames correctly
The database, cache, session and queue of the application must be configured with correct hostnames instead of localhost or 127.0.0.1. This way the services will be available across multiple hypernodes when the app is autoscaled.
Make sure the env variables (db, cache, session, queue) are not using localhost.
Application Specific Requirements - Magento 2
Supported CMS
Horizontal autoscaling is available for Magento 2.4.7 and higher. To make use of Horizontal autoscaling, there are a couple of other requirements the application should meet.
Make sure Redis cache is configured
Having a centralized cache database is essential for a consistent experience for your application.
Please make sure Redis cache is configured properly in the Magento 2 configuration file at <magento_root>/app/etc/env.php.
For the Redis hostname, use redismaster instead of localhost or 127.0.0.1.
More information about Redis on Hypernode.
Enable and configure Redis sessions
Redis sessions on the persistent instance is also required to make use of Horizontal autoscaling.
The persistent instance will be used to store the sessions so we can access the same sessions from the Horizontal autoscale Hypernodes.
You can check if Redis Persistent is enabled on your Hypernode by running
app@abcdef-example-magweb-cmbl:~$ hypernode-systemctl settings redis_persistent_instance
redis_persistent_instance is set to value True
If Redis Persistent instance is not enabled, you can enable the second Redis instance for sessions you run the command:
app@abcdef-example-magweb-cmbl:~$ hypernode-systemctl settings redis_persistent_instance True
Make sure Redis session is configured as described in our docs
Please notice the Redis host in the setup documentation. The Redis host should be set to redismaster instead of localhost or 127.0.0.1.
Make sure Elasticsearch/Opensearch is configured properly
Please make sure Elasticsearch or Opensearch host is set to elasticsearchmaster your Magento 2 configuration.
To set the hostname, you can run the following command for OpenSearch:
app@abcdef-example-magweb-cmbl:~$ bin/magento config:set catalog/search/opensearch_server_hostname elasticsearchmaster
For ElasticSearch:
app@abcdef-example-magweb-cmbl:~$ bin/magento config:set catalog/search/elasticsearch7_server_hostname elasticsearchmaster
More information about Elasticsearch on Hypernode.
Make sure RabbitMQ configured properly
Please make sure RabbitMQ host is set to rabbitmqmaster in the Magento 2 configuration file at <magento_root>/app/etc/env.php
More information about RabbitMQ on Hypernode
There is a rabbitmq user provisioned by Hypernode called hypernode-admin as a non-default user, but you can also configure RabbitMQ with a new different user of your own.
Please make sure to configure RabbitMQ without the default guest user.
For example, you can run the following command to change your RabbitMQ config in Magento 2:
bin/magento setup:config:set \
--amqp-host="rabbitmqmaster" \
--amqp-user="my_rabbitmq_user" \
--amqp-password="my_rabbitmq_password"
Make sure database storage is disabled & remote storage is enabled and configured
Remote storage is a requirement for Horizontal autoscaling. This is because the media files should be stored on a remote storage location instead of the local filesystem. This way the media files are available across multiple Hypernodes when the app is scaled up, and after scaling down the media files are still available.
Follow our documentation on remote storage for Magento 2 to configure remote storage.
Enabling Horizontal Autoscaling
For detailed steps on enabling horizontal Autoscaling, please refer to our documentation: How to enable horizontal Autoscaling?
Horizontal Autoscaling Pricing
To know more about the pricing, you can visit our documentation: Horizontal Autoscaling Pricing.