Statements Page

On this page

The Statements page of the DB Console helps you:

• Identify frequently executed or high latency SQL statements.
• View SQL statement details.
• Download SQL statement diagnostics for troubleshooting.

To view this page, click SQL Activity in the left-hand navigation of the DB Console. The Statements tab is selected.

Search and filter

By default, the Statements page shows SQL statements from all applications and databases running on the cluster.

Search statements

To search using the search field, type a string over Search Statements and press Enter. The list of statements is filtered by the string.

Date range

To search by date, click the date range selector and pick a date range that is within the time period. Click reset time to reset the date to the last hour.

It's possible to select a date range for which no statement statistics exist. CockroachDB persists statement statistics up to 1 million rows before the oldest row is deleted. The retention period of statistics is reduced the more active a workload is and the more distinct statement fingerprints there are.

Filter

To filter the statements, click the Filters field.

To filter by application, select App and choose one or more applications. When no application is selected internal statements are not displayed.

You can also filter by one or more databases (Database), SQL statement types (Statement Type), and for statement fingerprints that take longer than a specified time to run. To display only statements with queries that cause full table scans, click Only show statements that contain queries with full table scans.

The following screenshot shows the statements that contain the string rides for the movr application:

Statement statistics

Statistics aggregation is controlled by the aggregation interval property, set to 1 hour. Statistics between two hourly intervals belong to the nearest hour rounded down. For example, a statement execution ending at 1:50 would have its statistics aggregated in the 1:00-2:00 interval.

Aggregated statistics are flushed from memory to statistics tables in the crdb_internal system catalog every 10 minutes. The default retention period of the statistics tables is based on the number of rows up to 1 million records. When this threshold is reached, the oldest records are deleted.

To reset SQL statistics in the DB Console UI and crdb_internal system catalog, click clear SQL stats.

Example

See View historical statement statistics and the sampled logical plan per fingerprint.

SQL statement fingerprints

The Statements page displays SQL statement fingerprints.

A statement fingerprint represents one or more SQL statements by replacing literal values (e.g., numbers and strings) with underscores (_). This can help you quickly identify frequently executed SQL statements and their latencies.

For multiple SQL statements to be represented by a fingerprint, they must be identical aside from their literal values and placeholders.

These SQL statements:

• INSERT INTO new_order(product_id, customer_id, transaction_id) VALUES (380, 11, 11098)
• INSERT INTO new_order(product_id, customer_id, transaction_id) VALUES (192, 891, 20)
• INSERT INTO new_order(product_id, customer_id, transaction_id) VALUES (784, 452, 78)

have the fingerprint INSERT INTO new_order(product_id, customer_id, transaction_id) VALUES (_, _, _)

These SQL statements:

• INSERT INTO new_order(product_id, customer_id, transaction_id) VALUES ($1, $2, 11098)
• INSERT INTO new_order(product_id, customer_id, transaction_id) VALUES ($3, $4, 300)

have the fingerprint INSERT INTO new_order(product_id, customer_id, transaction_id) VALUES ($1, $1, _).

The following statements are not represented by either fingerprint:

• INSERT INTO new_order(product_id, customer_id, transaction_id) VALUES ($1, 11, 11098)
• INSERT INTO new_order(product_id, customer_id, transaction_id) VALUES ($1, $2, $3)

It is possible to see the same fingerprint listed multiple times in the following scenarios:

• Statements with this fingerprint were executed by more than one application_name.
• Statements with this fingerprint were executed both successfully and unsuccessfully.

Statements table

Click to select the columns to display in the table.

The Statements table gives details for each SQL statement fingerprint:

Statement Details page

Click a SQL statement fingerprint to open Statement Details. For each statement fingerprint, the details include:

• Overview
• Diagnostics
• Explain plan
• Execution stats

Overview

The Overview section displays the SQL statement fingerprint and essential statistics:

Mean statement time is the cumulative time taken to execute statements with this fingerprint within the aggregation interval.

• Planning time is the cumulative time taken by the planner to create an execution plan for statements with this fingerprint within the specified time interval.
• Execution time is the cumulative time taken to execute statements with this fingerprint in the specified time interval.

Resource usage displays statistics about storage, memory, and network usage for the SQL statement fingerprint.

• Mean rows/bytes read displays the mean number of rows and bytes read from the storage layer for statements with this fingerprint within the aggregation interval.
• Mean rows written displays the mean number of rows written by statements with this fingerprint within the aggregation interval.
• Max memory usage displays the maximum memory used by a statement with this fingerprint at any time during its execution within the aggregation interval or specified time interval.
• Network usage displays the amount of data transferred over the network for statements with this fingerprint within the aggregation interval. If this value is 0, the statement was executed on a single node.
• Max scratch disk usage displays the maximum amount of data spilled to temporary storage on disk while executing statements with this fingerprint within the aggregation interval.

Statement details displays information about the execution of the statement.

• Interval start time represents the start time of the statistics aggregation interval for a statement. For example, if a statement is executed at 1:23PM it will fall in the 1:00PM - 2:00PM time interval.
• Nodes: the nodes on which the statements executed. Click the node ID to view node statistics.
  
  Nodes are not visible for CockroachDB Serverless clusters.
• Regions: the regions on which the statements executed.
  
  Regions are not visible for CockroachDB Serverless clusters.
• Database: the database on which the statements executed.
• App: the name specified by the application_name session setting.
• Failed?: whether the statement failed to execute.
• Used cost-based optimizer?: whether the execution used the cost-based optimizer.
• Distributed execution?: whether the execution was distributed.
• Vectorized execution?: whether the execution used the vectorized execution engine.
• Transaction type: the type of transaction (implicit or explicit).
• Last execution time: when the statement was last executed.

Execution counts displays execution statistics for the SQL statement fingerprint.

• First attempts: the cumulative number of first attempts at executing statements with this fingerprint within the aggregation interval.
• Total executions: the total number of executions of statements with this fingerprint. It is calculated as the sum of first attempts and retries.
• Retries: the cumulative number of retries of statements with this fingerprint within the aggregation interval.
• Max Retries: the highest number of retries of a single statement with this fingerprint within the aggregation interval. For example, if three statements with the same fingerprint had to be retried 0, 1, and 5 times, then the Max Retries value for the fingerprint is 5.

Diagnostics

The Diagnostics section allows you to activate and download diagnostics for the SQL statement fingerprint.

When you activate diagnostics for a fingerprint, CockroachDB waits for the next SQL query that matches this fingerprint to be run on any node. On the next match, information about the SQL statement is written to a diagnostics bundle that you can download. This bundle consists of statement traces in various formats (including a JSON file that can be imported to Jaeger), a physical query plan, execution statistics, and other information about the query. The bundle contents are identical to those produced by EXPLAIN ANALYZE (DEBUG). You can use the information collected in the bundle to diagnose problematic SQL statements, such as slow queries. We recommend that you share the diagnostics bundle with our support team, which can help you interpret the results.

Activate diagnostics collection and download bundles

To activate diagnostics collection, click the Activate diagnostics button.

A row with the activation time and collection status is added to the Statement diagnostics table.

The collection status values are:

• READY: indicates that the diagnostics have been collected. To download the diagnostics bundle, click Bundle (.zip).
• WAITING: indicates that a SQL statement matching the fingerprint has not yet been recorded.
• ERROR: indicates that the attempt at diagnostics collection failed.

View and download diagnostic bundles for all statement fingerprints

Although fingerprints are periodically cleared from the Statements page, all diagnostics bundles are preserved. To view and download diagnostic bundles for all statement fingerprints do one of the following:

• On the Diagnostics page for a statement fingerprint, click the All statement diagnostics link.

• Click Advanced Debug in the left-hand navigation and click Statement Diagnostics History.

Click Bundle (.zip) to download any diagnostics bundle.

Explain Plan

The Explain Plan section displays CockroachDB's statement plan for an explainable statement. You can use this information to optimize the query. For more information about plans, see EXPLAIN.

By default, the explain plan for each fingerprint is sampled every 5 minutes. You can change the interval with the sql.metrics.statement_details.plan_collection.period cluster setting. For example, to change the interval to 2 minutes, run the following SET CLUSTER SETTING command:

> SET CLUSTER SETTING sql.metrics.statement_details.plan_collection.period  = '2m0s';

Execution Stats

The Execution Stats section has three subsections:

• Execution Latency by Phase displays the service latency of statements matching this fingerprint, broken down by phase (parse, plan, run, overhead), as well as the overall service latency. Overhead comprises the statements that remain after subtracting parse, plan, and run latencies from the overall latency. These might include fetching table descriptors that were not cached, or other background tasks required to execute the query.

  Note:

  Service latency can be affected by network latency, which is displayed for your cluster on the Network Latency page.

• Other Execution Statistics displays the following statistics.

  Statistic	Description
  Rows Read	The number of rows read by the statement.
  Disk Bytes Read	The size of the data read by the statement.
  Rows Written	The number of rows written by the statement.

• Stats by Node provides a breakdown of the number of statements of the selected fingerprint per gateway node. You can use this table to determine whether, for example, you are executing queries on a node that is far from the data you are requesting (see Optimize Statement Performance).
  
  Stats by Node are not visible for CockroachDB Serverless clusters.

See also

• Troubleshoot Query Behavior
• Transaction retries
• Optimize Statement Performance
• Support Resources
• Raw Status Endpoints
• Transactions Page

Was this helpful?