Hologres:Serverless Computing resource usage and billing

Permission Requirements

To estimate Serverless Computing consumption and billing from slow query logs, you need specific permissions. The following content describes the permission rules and grant methods.

• Query the data scan volume across all databases in an instance

  • Method 1: Grant the Superuser permission using the following command:

    -- Replace "Alibaba Cloud account ID" with the actual username.
    -- For RAM users, prepend the account ID with "p4_".
    ALTER USER "Alibaba Cloud account ID" SUPERUSER;

  • Method 2: Add the user to the pg_read_all_stats group using the following commands:

    Note

    In addition to the Superuser permission, Hologres supports the pg_read_all_stats group for viewing the data scan volume across all databases. If you are a regular user who needs full log access, you must contact a Superuser for authorization and to be added to the group.

    GRANT pg_read_all_stats TO "Alibaba Cloud account ID";-- Standard PostgreSQL authorization model
    CALL spm_grant('pg_read_all_stats', 'Alibaba Cloud account ID');  -- SPM
    CALL slpm_grant('pg_read_all_stats', 'Alibaba Cloud account ID'); -- SLPM

• View the data scan volume for the current database

  • Enable the simple permission model (SPM) or the schema-level simple permission model (SLPM), and then add the user to the db_admin group. The db_admin role can view the data scan volume for the current database. You can use the following commands:

    CALL spm_grant('<db_name>_admin', 'Alibaba Cloud account ID');  -- SPM
    CALL slpm_grant('<db_name>.admin', 'Alibaba Cloud account ID'); -- SLPM

  • Regular users can query their own data scan volume within databases associated with their account.

Query Resource Consumption for a Single SQL Statement

You can use the following SQL command to retrieve the detailed resource consumption for Serverless Computing jobs.

SELECT
    *,
    queue_time_ms,	-- Time the SQL spent waiting in the Serverless Computing queue, in milliseconds (ms)
    serverless_allocated_cores,	-- Number of CUs allocated by Serverless Computing to this SQL job
    serverless_resource_used_time_ms, -- Duration the SQL actually used Serverless Computing resources, in milliseconds (ms)
    (serverless_allocated_cores::DECIMAL(38, 4)) * (serverless_resource_used_time_ms::DECIMAL(38, 4)) AS serverless_cums
FROM
    hologres.hg_serverless_computing_query_log;

Query Resource Consumption Over a Specified Time Period

You can use the following SQL command to calculate the total resource consumption for successfully executed Serverless Computing jobs within a specific time window.

SELECT
    (SUM((serverless_allocated_cores::DECIMAL(38, 4)) * (serverless_resource_used_time_ms::DECIMAL(38, 4))) / 1000 / 60 / 60)::bigint AS serverless_cuh
FROM
    hologres.hg_serverless_computing_query_log
WHERE
    status = 'SUCCESS'
    AND serverless_allocated_cores IS NOT NULL
    AND serverless_resource_used_time_ms IS NOT NULL
    AND query_end BETWEEN '2024-05-01 00:00:00' AND '2024-05-31 24:00:00';-- Specify time range

Daily Usage Limit

Starting from Hologres V3.1.5, you can limit the daily Serverless Computing usage using the following SQL commands. Only Superusers can set these limits.

-- Set limit for a database
ALTER DATABASE <db_name> SET hg_serverless_computing_daily_max_cuh_usage_threshold = xx;

-- Set limit for a user
ALTER USER <user_name> SET hg_serverless_computing_daily_max_cuh_usage_threshold = xx;

• Limit Details

  • The default value is -1, which indicates that no limit is set.

  • The unit is CU·H.

  • Hologres aggregates and updates the total Serverless Computing usage for the instance every 10 minutes. Therefore, the enforcement of the limit may lag by up to 10 minutes.

  • Before executing a SQL statement, Hologres compares the current daily Serverless usage of the instance with the limit configured for the user or database. If the limit is reached, the SQL statement automatically falls back to using the compute resources of the instance. Otherwise, the SQL statement runs on Serverless resources.

  • Hologres tracks the daily Serverless usage based on the default time zone of the database. You can run the following commands in the same session to check the time zone of the current database:

    -- Reset time zone to avoid influence from client or user settings
    RESET timezone;
    
    -- Show the database's default time zone
    SHOW timezone;

• Recommendations

  • If you regularly use Hologres Serverless Computing, you can determine an appropriate daily limit by reviewing the usage information described in Query Resource Consumption Over a Specified Time Period.

  • If you are new to Serverless Computing, we recommend that you wait until your workload stabilizes before you set a daily usage limit.

To prevent queries from automatically falling back to instance resources when the limit is exceeded and instead cause the queries to fail immediately, you can run the following commands. Only Superusers can set this parameter.

-- Set at database level
ALTER DATABASE <db_name> SET hg_serverless_computing_enable_fallback_when_exceed_cuh_threshold = false;

-- Set at user level
ALTER USER <user_name> SET hg_serverless_computing_enable_fallback_when_exceed_cuh_threshold = false;

Cost Analysis

You can use the Cost > Cost Analysis module in Expenses and Costs to monitor and analyze the billing of Hologres compute resources.

1. Log on to Expenses and Costs. In the navigation pane on the left, click Cost > Cost Analysis.

2. On the Cost Analysis page, in the Condition List > Classification Dimension section on the right, select Billing Item. Then, select a Cost Type and Time Granularity as needed.

3. In the Filter Conditions section on the right, select the appropriate Billing Item based on your resource type.

For more information, see Cost Analysis.

Spending Alerts

You can use the Expenses and Costs service provided by Alibaba Cloud to manage budgets and set alerts for pay-as-you-go spending on specific products in specific regions. For more information, see Expense Alerts.

Single SQL Execution Duration Alert

Hologres supports monitoring for Serverless Computing metrics. We recommend that you create alert rules based on your business scenario to avoid unexpected charges.

For example, for the metric Longest duration among running Serverless Computing queries, we recommend the following alert rule:

Warning: The longest duration among running Serverless Computing queries is ≥ 3,600,000 milliseconds for 5 consecutive periods (1 period = 1 minute).

For more information, see Monitoring and Alerting Best Practices.