Displaying present location in the site.

HA Cluster Introduction Part 4: How to continue operations by failover

EXPRESSCLUSTER Official Blog

December 1st, 2021

Machine translation is used partially for this article. See the Japanese version for the original article.

Introduction

This time, I will explain "How other servers take over the operation of the application" of the HA cluster.

Applications and data

The key point of failover is that the applications on the standby server use the data taken over from the applications on the active server. Instead of simply stopping the application on the active server and starting the application on the standby server. Otherwise, if the application switches to a standby server due to failover, the data may be different.
Therefore, in HA clusters, the data used by the application is stored and shared in external storage for common access by each server. Applications on both the active and standby servers use the same data so that operations can be successfully taken over.

Client access destination

If a failover occurs while the client is accessing the IP address or host name of the active server, the client recognizes that the server is still down, even though the application is actually running on the standby server.
This is because the client continues to access the same server even if the application switches from the active server to the standby server.

To prevent this, HA clusters generally use virtual IP addresses. A virtual IP address is provided to the active server, and the client accesses the virtual IP address. In a failover situation, the virtual IP address provided to the active server is deleted and the virtual IP address is re-provided to the standby server. As a result, clients will be able to switch their access destination to the standby server.

Before and after failover occurred

Before you learn how to take over a failover, let's take a look at what happens before the failover occurs.
First, the data is stored in external storage, and the application on the active server is runnning using the data. In addition, the virtual IP address is assigned to the active server, and the client accesses the virtual IP address. On the other hands, the application on the standby server is in a stopped state and cannot access data from external storage (external storage is in an unmounted state).

When a failure occurs on the active server, "Application take over due to failover" will be executed.
First, the application on the active server will be stopped and the access to external storage will also be stopped (in an unmounted state). In addition, the virtual IP address assigned to the active server will also be deleted. At this point, both servers are in a "stopped state", "the access to external storage is also in a stopped state", and "virtual IP address is not assigned", and the system is down.

Next, a virtual IP address is assgined on the standby server and the application is started using data from external storage. At this point, the client can access the application on the standby server via the virtual IP address, and the failover operation will be completed.

Conclution

This time, We explained the mechanism of "Application take over due to failover".
In this blog, we are always looking forward to your requests.
If you have any questions about HA clusters or requests such as verification configurations, please do not hesitate to contact us if you have any questions.

Related Blog Posts