Hologres:Multi-cluster and auto scaling (Beta)

Multi-cluster

• Ideal use cases: high-concurrency, small-to-medium queries. Multi-cluster excels here by leveraging cluster isolation and FE Node load balancing. It supports more concurrent requests and automatically isolates workloads for better performance.

• Not suitable for: large, low-concurrency tasks. Scenarios like large-scale batch data ingestion (virtual warehouse 2 in How it works) are better suited for scale-up capabilities. Consider manual or scheduled scaling to allocate more resources to a single, powerful cluster.

Auto Scaling

Ideal use cases:

• High-concurrency, small-to-medium queries: See ideal use cases of Multi-cluster.

• Unpredictable workload peaks: This demand cannot be effectively managed by manual cluster increases or scheduled scaling.

Procedure

1. Log on to the Hologres management console. In the top menu bar, select a region.

2. In the left menu, choose Instances. Click your target instance ID to go to the details page.

3. In the left submenu, click Virtual Warehouse Management. On the page that appears, click the Auto-scaling tab.

4. Switch on Enable Auto Scaling. Set Maximum Clusters and click Save.

Example

After enabling Auto Scaling as shown above (with a single-cluster specification of 32 CUs, 1 reserved cluster, and a maximum of 4 clusters), verify its functionality as follows. This example uses pgbench, the native performance stress testing tool for PostgreSQL.

1. Create test tables in Hologres and load data.

   CREATE TABLE tbl_1 (col1 int, col2 int, col3 text);
   CREATE TABLE tbl_2 (col1 int, col2 int, col3 text);
   INSERT INTO tbl_1 SELECT i, i+1, md5(random()::text) FROM generate_series (0, 500000) AS i;
   INSERT INTO tbl_2 SELECT i, i+1, md5(random()::text) FROM generate_series (0, 500000) AS i;

2. On the stress testing server, create a SQL file named select.sql and add the following SQL statement:

   EXPLAIN ANALYZE SELECT * FROM tbl_1 LEFT JOIN tbl_2 ON tbl_1.col3 = tbl_2.col3 ORDER BY 1;

3. On the stress testing server, set the password as an environment variable.

   export PGPASSWORD='<AccessKey_Secret>'

4. Run the following stress testing command. For details, see Connect to a Hologres instance for data development.

   pgbench
   -c 30 \
   -j 30 \
   -f select.sql \
   -d <Database> \
   -U <AccessKey_ID> \
   -h <Endpoint> \
   -p <Port> \
   -T 1800

   During the stress test, the monitoring metrics for the virtual warehouse are as follows:

   • Cluster CPU utilization:

     • The load on the reserved cluster remains high, triggering Auto Scaling (Point 1) to add a new cluster.

     • After the test completes, the load on both clusters becomes low, triggering Auto Scaling (Point 2) to remove one cluster.

   • Virtual warehouse CPU utilization:

     • Before Auto Scaling adds a new cluster, the CPU utilization of the virtual warehouse consistently exceeds 85%.

     • After the new cluster is added, the overall CPU utilization of the virtual warehouse drops to around 70%.

Metrics

View the following metrics in the Hologres management console, and configure alert rules for them as needed. For details, see Monitoring metrics in the Hologres console.

• Cluster CPU utilization

• Cluster memory usage

• Number of cores provisioned by Auto Scaling for the virtual warehouse

Elastic event execution logs

1. Go to the Virtual Warehouse Management page and click the Elastic Event Execution Logs tab.

2. Select a time range to view the execution history of scaling events, which includes the execution time, virtual warehouse, execution status, event type, number of reserved clusters, and target number of clusters.

CloudMonitor events

CloudMonitor records all horizontal scaling events.

1. Go to the CloudMonitor Event Center. On the System Events page, select Hologres in the Event Monitoring area to monitor auto scaling events.

   • Instance:Warehouse:AutoElastic:Start: The start event for Auto Scaling in a virtual warehouse.

   • Instance:Warehouse:AutoElastic:Finish: The completion event for Auto Scaling in a virtual warehouse.

   • Instance:Warehouse:AutoElastic:Failed: The failure event for Auto Scaling in a virtual warehouse.

2. You can use these CloudMonitor events to configure notifications, alerts, and other actions. For details, see Using system events for alerting.

The following is an example of the event details for a failed Auto Scaling event:

{
    "status": "Failed",
    "instanceName": "<instance_id>",
    "resourceId": "<instance_resource_id>",
    "content": {
        "AutoElasticCPU": <cpu_num>,
        "ScaleType": "ScaleOut",
        "ScheduleId": "xxxxxx",
        "WarehouseId": "<warehouse_id>",
        "WarehouseName": "<warehouse_name>" 
    },
    "product": "hologres",
    "time": 1722852008000,
    "level": "WARN",
    "regionId": "<region>",
    "id": "<event_id>",
    "groupId": "0",
    "name": "Instance:Warehouse:TimedElastic:Failed"
}

ActionTrail

ActionTrail records operations performed in the Hologres management console, such as editing Auto Scaling settings, and the actual cluster scaling operations that Auto Scaling performs. For details, see Event audit logs.