You can upgrade your Elasticsearch (ES) cluster configuration if the existing setup cannot meet your business and performance requirements. This may be necessary when your data volume increases or memory usage becomes too high. During the upgrade, your cluster can remain available if the cluster load is low and your indexes have replicas. However, performance may be degraded and the upgrade may take a long time.
PrerequisitesEvaluate the specifications and capacity of your cluster. For more information, see Evaluate specifications and capacity.
Run the following command in the Kibana console to check for indexes in the close state. If any exist, you must change their state to open. Otherwise, the upgrade fails.
View index status
GET /_cat/indices?vSet the status of a closed index to open
POST /<index_name>/_openReplace <index_name> with the name of the index that is in the close state.
Run the following command in the Kibana console to check the number of replicas for your indexes. The number of replicas for each index must be at least 1. For multi-zone instances, the number of replicas for an index must be less than the number of zones. Before you upgrade, set the number of replicas to 1. After the upgrade is complete, you can manually increase the number of replicas as needed. For more information, see Index templates.
If the number of replicas is 0, the high-availability mechanism cannot be used, and service stability cannot be guaranteed.
GET _cat/indices?vNotes
Description
Service impact
Upgrading a cluster triggers a restart. The restart duration depends on factors such as cluster specifications, data structure, and data size. Perform this operation during off-peak hours. The restart usually takes a long time, often on the order of hours. For more information, see Restart an instance or a node.
ImportantIf the cluster indexes have replica shards and the cluster load is normal (CPU utilization is around 60%, heap memory usage is around 50%, and load_1m is lower than the number of CPU cores), the cluster can generally continue to provide services during the change.
If the cluster load is high and the indexes have no replicas, and there are many write or query operations during the upgrade or downgrade, your business may experience occasional access timeouts. Before changing the cluster, configure a client-side retry mechanism and increase the number of replicas to reduce the impact on your business.
Data is not lost during an upgrade. However, if the cluster load is high or the status is abnormal, you may experience occasional access timeouts or jitter. Configure a retry mechanism for your business and configure replicas for your index shards.
If the instance is in an abnormal state (yellow or red), select Force Change, Ignore Cluster Health Status when you upgrade the cluster. This operation may affect your services.
If you use a blue-green deployment for the upgrade, the node IP addresses change. If you specified node IP addresses in the cluster configuration, update them promptly after the change is complete.
Billing changes
After you submit an upgrade order, the instance is billed based on the new configuration. For billing rules, see Pay-as-you-go and Subscription.
NoteDuring the cluster upgrade, you can view the real-time cost of the new order on the Upgrade/Downgrade page.
Version upgrades
You cannot upgrade the version when you upgrade the cluster configuration. To upgrade the version, use the version upgrade feature.
NoteIf the version change does not meet the requirements during the cluster upgrade, the system reports an UpgradeVersionMustFromConsole error.
Since May 2021, Alibaba Cloud Elasticsearch no longer supports purchasing new instances of version 7.4. This does not affect existing instances. We recommend that you purchase a version 7.10 instance and then migrate your data.
Specification upgrades
You can change the specifications for only one node type at a time. Node types include data nodes, cold data nodes, client nodes, dedicated master nodes, Kibana nodes, and elastic nodes.
NoteTo reduce the number of data nodes, use the scale-in data nodes feature.
Disk type upgrades
You can upgrade disks with lower storage performance to disks with higher storage performance. The disk storage performance, from lowest to highest, is: ultra disk, standard SSD, and enterprise SSD (ESSD). You can upgrade as needed.
NoteThe system does not support upgrading from standard SSDs to ESSD-PL0.
Zone specification inventory
If you encounter insufficient inventory for the required specifications in a zone during a cluster upgrade, you can migrate the nodes in that zone before upgrading. For more information, see Migrate and upgrade a zone.
Shard allocation
After the change, shards are automatically balanced. In some scenarios, the load may be unbalanced. For example, improper shard settings can cause uneven data distribution across nodes. For information about how to troubleshoot and resolve unbalanced cluster loads, see Analysis and solutions for unbalanced cluster load issues.
ProcedureLog on to the Alibaba Cloud Elasticsearch console.
In the left-side navigation pane, click Elasticsearch Clusters.
In the top menu bar, select a resource group and a region.
In the instance list, find the target instance and click Upgrade in the right-side column.
On the upgrade page, click the node type that you want to upgrade and configure the parameters as required.
NoteYou can view the current configuration of the instance in the Current Configuration section.
On the upgrade page, you can modify configurations such as node specifications, storage type, and storage space per node. The available configuration items vary by node type. Refer to the page for the actual options.
For more information about instance configuration parameters, see Parameters on the buy page.
Modify the instance configuration as required. The following table describes some of the parameters.
Parameter
Description
Data Node
The maximum storage space of a single data node varies based on its disk type. You can view the specific limits in the console.
ImportantIf the cluster is in an unhealthy state (yellow or red), you cannot upgrade an ultra disk to more than 2 TiB. If you need more than 2 TiB, ensure that the cluster is in a healthy state, or increase the number of data nodes during the upgrade.
Instances with 1 CPU core and 2 GB of memory are suitable only for testing scenarios and not for production environments. According to the product Service-Level Agreement (SLA), they are not covered by after-sales support. We recommend that you upgrade instances with 1 CPU core and 2 GB of memory to a higher specification as soon as possible.
Dedicated Master Node
You can purchase new dedicated master nodes or upgrade the specifications of existing ones.
Cold Data Node
You can purchase new cold data nodes or upgrade the specifications of existing ones.
Client Node
You can purchase new client nodes or upgrade the specifications of existing ones.
Kibana Node
You can upgrade the specifications of an existing Kibana node.
ImportantWhen you purchase an Alibaba Cloud Elasticsearch instance, the Kibana node is enabled by default and cannot be changed. Due to performance and stability considerations, we recommend purchasing a Kibana node with 2 CPU cores and 4 GB of memory or higher. A Kibana node with 1 CPU core and 2 GB of memory is provided free of charge but is recommended only for testing scenarios.
Force Change
If the instance is in an abnormal state (red or yellow) and services are severely affected, and you need to scale out immediately to resolve the issue, we recommend using Force Change.
ImportantForce Change ignores the instance health status and can speed up the change process. However, the actual time required for the cluster to recover depends on the data volume, cluster pressure, and configuration.
Force Change restarts the instance, which may cause service instability during the restart phase.
If Force Change is not enabled, the system uses the Restart method to upgrade the cluster by default. For related precautions, see Restart an instance or a node.
Intelligent Change
Enabled by default. When Intelligent Change is enabled, the backend automatically selects a suitable change method based on your operation type. You can also manually disable Intelligent Change. After disabling it, you can specify a Change Method. The supported methods are Blue-Green Deployment and In-place Change:
Blue-Green Deployment: Adds the same number of new nodes to the cluster and copies data, then seamlessly switches to the new nodes. This process is relatively smooth but takes longer. A blue-green deployment causes the node IP addresses to change. If you specified node IP addresses in the cluster configuration, update them promptly after the change is complete.
In-place Change: Performs a rolling change on the required nodes in the cluster one by one. This change does not require data copying, and the time required is not affected by the data size, but it will have a certain impact on the cluster's performance.
For more information, see Introduction to change methods.
NoteIf you have no special business requirements, we recommend that you enable Intelligent Change.
Read and select the terms of service and click Buy Now.
After you complete the payment, the cluster restarts to apply the new configuration. The upgrade is complete when the cluster has finished restarting.
Does Alibaba Cloud ES support version upgrades or downgrades?
After changing the number of nodes, does the cluster automatically rebalance shards?
Does changing the cluster configuration affect the ES service?
What do I do if I selected the wrong configuration when purchasing an ES instance?
After upgrading the instance type, can I downgrade the configuration? How?
How can I change the cluster configuration to handle a temporary surge in traffic?
What do I do if the "UpgradeVersionMustFromConsole" error is reported when I upgrade a cluster?
Under what circumstances can I use the force restart feature for ES, and what are the effects?
What do I do if an error or timeout occurs when upgrading a cluster?
Will changing the disk type of an ES instance cause data loss?
Can I upgrade the CPU and memory of an ES instance without data migration?
API documentation for upgrading a cluster: UpdateInstance
To view the change progress and status, see View change progress and change records.
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4