OpenSearch
OpenSearch provides full-text search and analytics as a project service. The initial offering uses three identical persistent nodes. Availability depends on the project's infrastructure and operator qualification; trial organizations cannot provision it.
Create a cluster
Open Services → Create Service → OpenSearch. The creation form is organised into five sections:
- Basic Configuration — cluster name and OpenSearch version. Names ending in
-nodes,-discovery, or-dashboardsare reserved and rejected. - Resources — compute class (
ecofor economic shared vCPU,standardfor balanced shared vCPU, orproductionfor dedicated vCPU), flavor (vCPU and memory per node), storage size in GiB per node, and storage class (standardorperformance). The storage class is fixed at creation. The compute class can be changed after creation; changing it reschedules the nodes one at a time. The flavor and storage size can only grow. - OpenSearch API — optional public HTTPS access with an allow-list of CIDR blocks.
- Dashboards — optional OpenSearch Dashboards pod with its own public access and CIDR allow-list.
- Backups — scheduled snapshot configuration. Enabled by default with a daily run at 02:00 UTC. You can choose which days of the week to run and at what hour (UTC). Disable to stop scheduled snapshots; manual snapshots remain available either way.
Each node has at least 1 vCPU, 2 GiB RAM, and 20 GiB storage. A minimum cluster therefore allocates 3 vCPU, 6 GiB RAM, and 60 GiB of storage.
The three nodes run on different workers. Each participates in cluster management, storage, and ingestion. The initial offering has no single-node option and no horizontal resizing.
Secure access
The OpenSearch endpoint uses private HTTPS with authentication by default. Use the CA certificate supplied with the credentials to verify the server; do not disable certificate verification.
Public API access is optional. Enable it in the service configuration and, if needed, restrict it to IP addresses or CIDR ranges. An empty allow-list permits all client addresses, while authentication remains required.
For applications, create a connection from the container or job, select OpenSearch, and specify its allowed index patterns, such as products-*. Each connection receives a separate user with the chosen data permissions and trusted CA:
- Read-only — read and search matching indices.
- Read/write — read, write, create indices and manage their mappings within the allowed patterns.
- Full — all index operations on matching indices, but no cluster or security administration.
Cluster credentials are available separately to developers for managing data and Dashboards. They grant full index operations plus content-administration features: index templates, ingest pipelines, search pipelines, stored scripts, ISM policies, alerting, anomaly detection, asynchronous search, point-in-time queries, and observability objects. Security configuration, snapshot repositories, cluster settings, ML connectors, and notification channels remain platform-managed. Use scoped credentials for applications.
The cluster password is returned only once, when you first reveal the credentials. Store it immediately. A later read returns the username, CA certificate, and endpoint but omits the password. To obtain a new password, rotate the credentials.
Dashboards
Dashboards is disabled initially. Enable it when you need the OpenSearch web interface. Its public HTTPS access and address restrictions are configured independently: publishing Dashboards does not publish the OpenSearch API. Always use the Dashboards endpoint returned by the SailOps UI or API; do not construct the address yourself. SailOps may truncate and hash long hostname labels. Dashboards is included with the cluster at no extra cost.
Snapshots and restoration
S3 snapshots run on a configurable schedule (daily at 02:00 UTC by default) with seven-day retention. You can change the days and hour, or disable scheduled snapshots entirely, from the Backups section of the service form or via the API. The Backups tab lists snapshots and lets developers request an additional one or delete one; individual S3 objects must never be deleted because incremental snapshots share data.
Select a completed snapshot and Restore. Enter a new cluster name. Restoration provisions a new three-node instance, leaves the source intact, and reads the source snapshot repository without writing to it. Security is recreated from the managed service configuration rather than importing the source security index. Snapshots exclude the security index, global cluster state, and the management index containing snapshot and Index State Management (ISM) policies. The service recreates its backup policy; recreate any custom ISM policies separately. Dashboards saved objects are not imported into the new instance; export and import them separately. Reconnect applications to the new instance after checking its data.
Monitoring and costs
The service overview shows a unified status and CPU/memory/storage usage. The possible statuses are Running (all nodes serving requests), Updating (a rolling operation such as a compute-class change is in progress), Warning (serving requests with unassigned replica shards), Error (at least one primary shard unavailable), Pending (provisioning or starting), Stopped (compute suspended, retained storage still billable), and Deleting (the cluster is being deleted).
The Observability tab displays historical CPU, memory, and storage utilization charts for the cluster. It uses the same dashboard as other services, with configurable time ranges and auto-refresh. The Logs tab provides cluster logs.
The cluster is billed on its nodes and their storage; Dashboards is included. Provisioned PVC capacity remains billable while a project is suspended; compute stops only after suspension has completed. Snapshot storage is measured as unique repository bytes, rather than the sum of logical snapshot sizes.