Alibaba Cloud optimizes the kernel of ApsaraDB for Redis in a continuous way to fix
security vulnerabilities and improve service stability. You can upgrade the kernel
version or the minor version of an ApsaraDB for Redis instance with a few clicks in
the console.
Procedure
The upgrade procedure varies by architecture. For more information, see the following
table.
Apply for resources that are required to create an instance of a new version, such
as proxy node resources.
Synchronize the full and incremental data from the original instance to the instances
of the new version.
Switch over instances. When data synchronization is close to completion, the original
instance is set to read-only and waits less than 60 seconds for data to be completely
synchronized. After data synchronization is complete, the system disassociates the
virtual IP address (VIP) from the original instance and associates the VIP with the
new instance.
Note If you select Update During Maintenance, the system performs the switchover operation within the maintenance window of the
instance.
Complete the upgrade. Release the resources of the original instance and change the
state of the new instance to Running.
Upgrade the replica node. Stop the original replica node and then start a replica
node of the new version.
Synchronize data from the master node to the replica node of the new version.
Switch over the master node and the replica node. When data synchronization is close
to completion, the instance is set to read-only and waits less than 60 seconds for
data to be completely synchronized. After data synchronization is complete, the system
disassociates the VIP from the master node and associates the VIP to the replica node.
Meanwhile, the system performs a master-replica switchover to promote the replica
node of the new version to the master node.
Note If you select Update During Maintenance, the system performs the master-replica switchover operation within the maintenance
window of the instance.
Upgrade the master node that is demoted to replica node. The original master node
becomes the replica node. The system repeats Step 1 and Step 2 to upgrade the replica
node and synchronize data.
Complete the upgrade. The status of the instance is changed to Running.
Impacts
The Redis service remains unaffected during the phases of applying for resources,
upgrading replica nodes, and synchronizing data.
During the instance switchover or master/replica switchover, the instance experiences
transient connection errors. The instance stays in the read-only state for 60 seconds
or less to wait until all data is synchronized. Upgrade the instance during off-peak
hours and ensure that your application is configured with a reconnection mechanism.
On the top of the page, select the region where the instance is deployed.
On the Instance List page, click the Instance ID of the instance.
In the Basic Information section, click Minor Version Upgrade.
Figure 1. Upgrade the minor version
Notice The system automatically detects the minor version of the instance. If you cannot
click Minor Version Upgrade, it indicates that the instance is upgraded to the latest minor version.
In the dialog box that appears, specify a time to upgrade the minor version.
Figure 2. Specify a time to perform the task.
Note The instance experiences transient connection errors within seconds and stays in the
read-only state for 60 seconds or less during instance switchover or master/replica
switchover. We recommend that you select Update During Maintenance. This way, the system performs the switchover operation within the maintenance window
of the instance to minimize the impact. For more information about how to modify the
maintenance window, see Set the maintenance window.
Click OK.
FAQ
Q: Why does the state of the instance change to Upgrading Minor Version after I select
Update During Maintenance to upgrade the minor version?
A: When the system prepares for upgrades, such as applying for resources and synchronizing
data, instances or master and replica nodes cannot be switched over. Therefore, the
system does not affect instances that provide services.
Note The instance experiences transient connection errors within seconds and stays in the
read-only state for 60 seconds or less only during the switchover of instances or
master and replica nodes.