OpenSearch - Get started

Deploy a cluster

Connect to ITCare, search for the Global Service to attach the cluster to and click on it.

Click Create resource on the left control panel and select OpenSearch.

Give the cluster a unique name and define a prefix name that will be used to name the virtual machines. Click Next.

Select number of nodes and node size. Click Next.

Select storage volume. Click Next.

Select the region and area to which you want to locate your cluster. Click Next.

Select VLAN you want to deploy your cluster into. Click Next.

Enable or disable additional options:

• Virtual machines and clusters monitoring

• 24/7 monitoring

• Backup

• Virtual machines replication (Disaster recovery)

Click Next.

Select the version and define an administrator password that will be used to manage your cluster.

Passwords are not saved by cegedim.cloud. Make sure to save your password.

Verify settings, here you can:

• Check virtual machine names to be created.

• Save your administrator password.

• Modify the management options.

Click Submit.

Once the cluster is ready, you will be notified by email with the information required to connect to the cluster.

Cluster creation can take up to 2 hours based on the current load on automation. The cluster will then be be displayed in your Global Service, in the left control panel, under the related cluster section. The green arrow indicates that the cluster is active.

Start a cluster

On the left control panel, click the name of the cluster. The cluster page is displayed. At the top of the cluster page, click the Manage button, then Start and confirm.

Starting a cluster will start all virtual machines attached to the cluster.

An email notification will be sent when the service will be activated.

Stop a cluster

At the top of the OpenSearch cluster page, click the Manage button, then Stop.

Input an RFC number for tracking (optional). Click on Submit.

Stopping a cluster will stop all virtual machines attached to the cluster and monitoring will be disabled.

An email notification will be sent when the cluster is stopped.

Add Nodes

At the top of the cluster page, click the Manage button, then Add Nodes.

Select the number of nodes you want to add (even number) and select the new size (cpu/ram). Specify the data disk size.

An email notification will be sent when all node are added.

Resize Nodes

At the top of the cluster page, click the Manage button, then Resize Nodes.

Select the nodes you want to resize and select the new size (cpu/ram)

Each node will be sequentially resized and restarted.

An email notification will be sent when all node are resized.

Delete a cluster

At the top of the cluster page, click the Manage button, then Delete.

This action will stop and delete all virtual machines. All CI will be removed and disappear from your Global service.

Please note that this action is not recoverable.

Input an RFC number for tracking (optional) then click Submit.

An email notification will be sent when the cluster is removed.

Upgrade a cluster

Upgrading a cluster is not yet implemented in ITCare.

If you want to upgrade your existing cluster to a newer available version you must send a request to the support indicating which version you would like.

During the upgrade the OpenSearch cluster will continue to work correctly as we're doing a rolling upgrade. But you might see some short indisponibility during the upgrade of the dashboard servers which is done after the upgrade of the Elastic cluster.

Upgrade the cluster topology

The cluster topology upgrade is proposed for an Opensearch cluster initially consisting of three nodes (basic topology).

The upgrade is implemented in ITCare via "Manage - Migrate to dedicated master" option.

Migration allows to add two Master nodes, while specialising the existing (n-1) nodes to the Data role. The cluster will thus be composed of three Master nodes and the rest of the nodes dedicated to the Data role.

During the upgrade to a "Dedicated Master" topology, the cluster will continue to work, although it may temporarily enter a "Yellow" status. However, some cluster objects, such as dashboards, may be temporarily unavailable. Everything will return to normal once the migration is complete.

Migration demo

Add ingest nodes to the cluster

The functionality to add dedicated ingestion nodes is available for an OpenSearch cluster with a "dedicated master" topology (five or more nodes).

The deployment of dedicated ingestion nodes is implemented in ITCare through the "Manage - Add Ingest Nodes" option.

With these dedicated ingestion nodes, it is possible to isolate the data ingestion process, which helps minimize the impact on indexing or search performance, even when dealing with large volumes of incoming data. This improves the overall stability and performance of the cluster.

Link to official documentation

Ingestion nodes demo

Set node attributes

Node Attributes - Multi-Tier Architecture - Hot/Cold

The multi-tier architecture (Hot-Warm-Cold) in OpenSearch allows for optimization of costs, performance, and scalability based on specific application needs. This architecture enables data storage based on access frequency, distributing frequently accessed data to identified nodes (Hot), moderately accessed data to identified nodes (Warm), and infrequently accessed data to identified nodes (Cold). In addition to attribute differences, the number of nodes at each level can also vary to meet specific search query needs.

At cegedim.cloud, to meet the needs of our customers, we offer two levels of the multi-tier architecture (Hot-Cold). Node attribute configuration is available in ITCare via the "Set node attributes" action, accessible at each node level.

For more information on node attribute configuration, please refer to the official documentation

Exceed Maximum Shard Limit

In Opensearch, each node has a default shard limit of 1000 shards.

This limit is in place to help maintian the performance and stability of your cluster by controlling resource usage per node. Exceeding this limit, such as when creating new indexes, will result in error.

If you encounter this error, consider the following actions below to resolve it.

Add More Data Nodes to the Cluster

Adding additional nodes can distribute the shards across more resources, reducing the load per node. Each additional node will bring its own 1000 shards limit, effectively increasing your cluster's overall shard capacity.

Increase cluster_max_shard_per_node Setting

Adjust the cluster.max_shards_per_node parameter to allow more shards per node if you have sufficient hardware resources (CPU, memory and storage). We recommend to not exceed 2000 shards per node to avoid overloading.

Use the following command to update the parameter:

PUT /_cluster/settings
{
    "persistent":
    {
        "cluster.max_shards_per_node":<desired_shard_limit>
    }
}

Optimize Shard Allocation per Index

When creating new indices, carefully plan the number of shards to match the index's data size and expected growth.

Avoid over-sharding by:

• Using fewer, larger shards for smaller datasets.

• Periodically reviewing shard allocation to ensure efficient resource usage.

Important Notes

You should regularly monitor shard and resource utilization in your cluster to ensure optimal performance.

A Grafana dashboard for OpenSearch is available for this effect in your Advanced Metrology platform.

You should ensure your nodes are equipped with sufficient resources before increasing shard limits.