Alibaba Cloud Elasticsearch provides two cluster update methods: in-place update and blue-green update. As your business grows, changes in your data volume and access traffic may require cluster update operations. These operations include scaling a cluster in or out, upgrading the cluster version, and migrating nodes to another zone.
In most cases, the system automatically selects an appropriate update method based on the type of update operation that you perform. When you restart a cluster or upgrade the configuration of a cluster, you can specify an update method based on your business requirements.
Considerations
Before you update an Alibaba Cloud Elasticsearch cluster, verify the following:
Replica shards — Each index in the cluster has at least one replica shard. Replica shards improve cluster stability during an update.
Multi-zone clusters — The number of replica shards per index is less than the number of zones in which the cluster is deployed. After the update completes, you can manually increase the number of replica shards.
How each update method works
In-place update
The system updates the nodes that require an update one at a time in a rolling manner. An in-place update does not copy data, so the update duration is not affected by the data volume.
An in-place update may affect cluster performance. Avoid performing the update during peak business hours.
Blue-green update
The system adds the same number of new nodes to the cluster and copies data to the new nodes. The cluster then switches over to the new nodes, and the system removes the original nodes. A blue-green update is relatively smooth but takes longer than an in-place update, and the node IP addresses change after the update completes.
The duration of a blue-green update depends on factors such as the cluster specifications, data structure, and data size. The update typically takes hours.
In extreme cases, read operations may be affected during a blue-green update. Perform the update during off-peak hours and configure a retry mechanism for your client.
Choose an update method
Some update operations support only one update method. Check the following constraints before you select a method for your operation.
Operations that require a blue-green update
Zone migration — A blue-green update is the only supported method when you migrate nodes to another zone.
Disk type change — When you upgrade the configuration of a cluster, you can specify an update method. If the update changes the disk type of nodes, you can specify only a blue-green update.
Ultra disk expansion beyond 2 TB — When you upgrade a cluster and change the ultra disk storage space to more than 2 TB, a blue-green update is the only supported method.
Operations that do not support a blue-green update
Forced restart — Only an in-place update is supported.
Client nodes only — If a cluster has client nodes but no dedicated master nodes, you cannot perform a blue-green update on the client nodes only.
Dedicated master nodes only — You cannot perform a blue-green update on the dedicated master nodes only.
Recommended method by scenario
If both update methods are available for your operation, select a method based on your scenario.
| Scenario | Recommended method |
| The cluster encounters performance bottlenecks and you want to complete the update quickly. | In-place update |
| The update duration is not critical, but high cluster availability is required. | Blue-green update |
| An individual node in the cluster delivers poor performance, for example, sustained high CPU utilization. | Blue-green update |
Next steps
To specify an update method when you restart a cluster, see Restart a cluster or node.
To specify an update method when you upgrade the configuration of a cluster, see Upgrade the configuration of a cluster.