Skip to content

Configure MCP Pooling

Use these procedures to enable a pooled Hydrolix MCP server deployment and override query routing for individual pools. For an overview of how pooled MCP tenancy works, see MCP Pools.

To set up a pooled MCP deployment:

  1. Set cluster-wide defaults with the mcp_hydrolix tunable.
  2. Create an additional pool with service: mcp-hydrolix.
  3. Optionally, override per-tenant query routing on that pool with a routing: block.

Enable the MCP server⚓︎

MCP is enabled by default through the mcp_hydrolix tunable. This example sets cluster-wide connection, worker, and transport defaults under spec.mcp_hydrolix.

Cluster-wide MCP Configuration
1
2
3
4
5
6
7
8
9
spec:
  mcp_hydrolix:
    enabled: true
    hydrolix_connection:
      query_pool: default-query-pool
      query_timeout_sec: 30
    mcp_server:
      transport: http
      bind_port: 8000

For the complete list of cluster-wide fields, see The mcp_hydrolix tunable.

Create an mcp-hydrolix pool⚓︎

Create a dedicated MCP pool by adding a pool definition to hydrolixcluster.yaml with the service field set to mcp-hydrolix.

MCP Pool
1
2
3
4
5
6
7
8
spec:
  pools:
    analytics-mcp:
      name: analytics-mcp
      service: mcp-hydrolix
      cpu: 2
      memory: 4Gi
      replicas: 2

Without a routing: block, the pool inherits every setting from the cluster-wide mcp_hydrolix tunable.

Override per-pool query routing⚓︎

Add a routing: block to the pool to override a subset of query-routing settings for that pool only. Any unset field inherits the cluster-wide mcp_hydrolix value.

Per-Pool Routing Override
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
spec:
  pools:
    analytics-mcp:
      name: analytics-mcp
      service: mcp-hydrolix
      cpu: 2
      memory: 4Gi
      replicas: 2
      routing:
        query_pool: analytics-query-pool
        query_timeout_sec: 60
        max_result_cells: 100000

See Per-pool routing overrides for the full list of overridable fields.

Apply the configuration⚓︎

Apply the updated cluster definition with kubectl.

Apply the Cluster Configuration
1
kubectl apply -f hydrolixcluster.yaml