Skip to content

Explorer


Database display covers data from a variety of infrastructure databases that DataKit supports collecting. Through a centralized Explorer, you can monitor key metrics of databases such as MySQL, SQL Server, PostgreSQL, Oracle, and MongoDB in real time, including version, uptime, load trend, blocking rate, and more.

Note

The DataKit collector has been updated. Contact your administrator to upgrade to the new version and enable the new database monitoring experience.

Data Display

The Database Explorer provides multiple professional analysis views based on lists and charts.

Displays the latest metric data for the corresponding databases in the current workspace, including database address, instance name, database type, version, uptime, QPS, load trend, blocking rate, and more.


Uses chart types such as Top List, pie chart, treemap, and grouped table chart to filter data under by conditions based on count, last, first, and count_distinct calculation modes.


Data List

The database list page displays the core status of each instance in card-style rows. The corresponding type icon is displayed before the database address, making it easy to quickly identify database types such as MySQL, PostgreSQL, Oracle, SQL Server, and MongoDB.

The list fields are described below:

Field Description
Database Address Connection address of the database instance; displayed in full on the details page without width restrictions
Instance Name Database instance identifier
Type Database type, such as MySQL, PostgreSQL, etc.
Version Database version number
Uptime Time the instance has been running
QPS Queries per second, including the current value and trend thumbnail
Load Trend Session load distribution trend aggregated by wait type
Blocking Rate Proportion of current database blocking

Time Widget

In the upper right corner of the Explorer, you can select data for different time ranges.

Data Details

In the Database Explorer, click a data record to open the corresponding details page in a side panel. You can view the record's Overview, Query, Sessions, Metrics, Logs, custom associated views, and other data in one place, avoiding repeated navigation and queries.

Overview

The Overview page displays the core running status of the database instance. The top of the page shows the instance name, version, uptime, slow query log status, and host information, helping you quickly locate the infrastructure resources where the database instance resides.

Execution Count / Total Execution Time

The left Y-axis uses a bar chart to display the total execution count of all queries within the time interval, and the right Y-axis uses a line chart to display the total execution time of all queries within the time interval.

Top 5 Queries

Displays the top queries sorted by total execution time, including query statement, execution count, average execution time, total execution time, and performance overhead percentage. Click any query to enter Query Overview and view details.

  • Sorted by total execution time in descending order by default;
  • Click any row to open the Query Overview side panel.

Databases (PostgreSQL)

On the Overview page of a PostgreSQL instance, the Databases table is located after Top 5 Queries and before Key Metrics. By default, it spans the full width and displays capacity, connection, transaction, data read, and maintenance metrics for collected databases in the current instance.

Each row represents a logical database in the current instance, such as postgres and loadtest. The table displays only databases within the collection scope of the current instance; the count indicates the number of collected databases and does not represent the total number of databases in the instance. If a field is missing, the row containing that database is still retained.

Top 5 Queries is used to identify higher-overhead SQL; the Databases table is used to compare the overall running status of each database. A transaction can contain multiple SQL statements, so the transaction commit rate and SQL execution count use different statistical bases.

Field Description

Field Metric Source Description Unit
Database db Name of the database within the collection scope of the current PostgreSQL instance. Text
Database Size database_size The latest valid database space usage within the selected time range. Raw values are processed as bytes and converted to capacity units for display. B, KB, MB, GB, TB
Active Connections connection_count Number of connections in the state=active state in the most recent complete valid sample, aggregating connections by user and application for the current database. count
Commit Rate xact_commit Based on the cumulative number of committed transactions for the database, calculates the average number of commits per second over valid sampling intervals within the selected time range. transactions/sec
Rollback Rate xact_rollback Based on the cumulative number of rolled-back transactions for the database, calculates the average number of rollbacks per second over valid sampling intervals within the selected time range. transactions/sec
Block Read Rate blks_read Uses the database-level statistics series to calculate the average number of blocks read per second over valid sampling intervals within the selected time range. This value is not directly equivalent to physical disk IOPS. blocks/sec
Buffer Hit Rate blks_hit Uses the database-level statistics series to calculate the average number of buffer hits per second over valid sampling intervals within the selected time range. It includes only PostgreSQL buffer hits. This value is a rate, not a hit ratio. blocks/sec
XID Wraparound Remaining wraparound The transaction ID wraparound remaining reported by the collector, taking the latest valid value. The smaller the value, the more attention should be paid to transaction freezing and VACUUM; the specific calculation boundary depends on the collector version used. transaction IDs
Autovacuum Cumulative Count autovacuum_count Takes the latest valid cumulative value for each user table within the collection scope, then summarizes by database; only automatic VACUUM operations are counted. times
Estimated Dead Tuples n_dead_tup Takes the latest valid estimated dead tuple value for each user table within the collection scope, then summarizes by database. Dead tuples are usually generated by UPDATE and DELETE and do not indicate data corruption. rows

Time Range and Calculation Basis

The table follows the time range in the upper right corner of the page:

  • Database Size, Active Connections, and XID Wraparound Remaining take the latest valid sample value within the selected time range and are not summed across time. Active Connections are aggregated within the same complete sample to avoid combining user and application groups from different times.
  • The raw metrics for commits, rollbacks, block reads, and buffer hits are cumulative values. The interval average rate is calculated as "sum of valid sample increments ÷ sum of corresponding valid sampling durations (seconds)"; it does not directly sum raw cumulative values, nor does it add up the rates at each point in time. For example, if the number of committed transactions increases from 824031 to 824057 within 10 seconds, the commit rate for that interval is (824057 - 824031) / 10 = 2.6 transactions/sec.
  • Autovacuum Cumulative Count and Estimated Dead Tuples first take the latest valid value for each collected user table, then sum by database; values for the same table at different sampling times are not accumulated repeatedly. Autovacuum Cumulative Count is not the number of new executions within the selected time period and may decrease due to statistics reset, table deletion, or changes in collection scope.

Data Scope and Missing Values

Collection Scope Affects Displayed Results

Table-level metrics are affected by the relations configuration and represent only statistics for collected user tables. When only some tables are collected, you can learn about data coverage from the collection scope hint. The collection scope of database-level metrics such as Database Size and table-level aggregated metrics may differ.

blks_read and blks_hit in collected data may also be used for SLRU statistics; the Databases table uses the database-level statistics series of the target database.

For configuration and permission requirements, refer to PostgreSQL Collector Configuration.

When collected and the result is confirmed to be zero, 0 is displayed. When data is not collected, no valid data exists, data is expired, or calculation is not possible, — is displayed, with a hint explaining the reason; when only part of the time range is covered, the actual data coverage range is indicated.

Counter resets or collection interruptions may affect rate calculation; when valid samples are insufficient, missing results must not be treated as zero. The absence of a returned state=active group does not necessarily mean that Active Connections is zero; determine this in conjunction with the collector's zero-value reporting rules.

Table Operations

  • By default, sorted by Database Size in descending order, with missing values at the end; numeric columns are sorted by raw numeric values.
  • Supports fuzzy search by database name.
  • All 10 columns are retained by default. On narrow screens, the database name column is fixed, and the remaining columns can be viewed by scrolling horizontally.
  • Hover over the hint icon in the table header to view metric meaning, unit, and calculation basis. Headers use single-line names, and the capacity unit for Database Size is displayed with the value.
  • Database names are displayed as plain text. To further analyze SQL, sessions, or table structures, use the Query, Sessions, and Tables tabs in the instance details.

Key Metrics

1. Query Wait State Distribution

Displays the wait event distribution of current database sessions as a donut chart, aggregated by wait state.

2. Load Trend Chart by Wait Type/Database/User

The bar chart displays database load distribution by wait type/database/user, visually presenting changes in the proportion of load types such as Idle, CPU, Application, and System I/O.

3. Key Resource Metrics

Displays core resource consumption trends for the node where the database resides, including:

  • CPU: CPU utilization, CPU load
  • Disk: disk read/write latency, bytes read/written per second, IOPS, disk usage
  • Network: network traffic/sec, network packets/sec
  • Memory: memory utilization, memory usage

Metrics

Displays all database performance metrics in visual charts, including core metrics such as session count, response time, cache hit ratio, tablespace usage, and PGA memory.

Query

Focuses on SQL performance analysis to help locate slow queries and performance bottlenecks.

The top displays load trend charts aggregated by three dimensions: wait type, database, and user; below is the query list, which displays the execution count, average execution time, total execution time, and performance overhead percentage of each SQL statement, and supports sorting by each dimension.

  • Click any query to open the Query Overview side panel and view the SQL statement, performance trend chart, and sampling records;
  • Click a sampling record to drill down to query details and view the complete log data for that sample;

  • In the upper right corner of the SQL statement area, click AI Optimization Suggestions to open the TOBY AI TruePilot side panel and automatically run optimization analysis for that query.

Sessions

Focuses on connection sessions, lock waits, and blocking analysis to help diagnose concurrency conflicts.

  • The top displays load trend charts aggregated by three dimensions: wait type, database, and user; the vertical axis represents the number of active sessions at the corresponding point in time;
  • Below is the snapshot list, which displays session data aggregated by collection time point, including key metrics such as current connection count, wait event distribution, number of waiting connections, involved users, and average execution duration.

    • Click a snapshot to expand and view details; after expansion, it displays a statistical summary of the root blocking session count and waiting query count, as well as a session details list;

    • Provides a "Show only blocking connections" filter toggle:

      • Upper right corner of the snapshot list: globally filters snapshots with blocking activity;
      • Upper right corner of the expanded snapshot details: displays only blocking-related sessions in the current snapshot.

Click a query statement in the session list to further drill down and view detailed SQL performance data.

Logs

View database-related log data, including slow query logs and error logs.

Tables

View database table structure information. Use the Database drop-down list to switch the target database, or search for a table name in the search box.

The list displays all tables in the current database, including table name, column count, primary key, engine, row format, and creation time. Click a table name to expand and view the field structure details of that table.

Extended Fields

Displays the complete metadata of the instance as key-value pairs, with support for searching field names or values.

Custom Views

Centralizes user-defined views and supports adding new views.