Monitoring

MatrixOne records every statement it runs in system.statement_info: the SQL text, duration, rows and bytes read, errors and, for slow statements, the execution plan. There is nothing to enable. MySQL's slow_query_log, long_query_time and general_log variables exist for compatibility, but MatrixOne's history lives in this table.

• Statement history
• Slow queries from the command line
• Execution plans
• Table statistics
• Server logs and metrics
• Configuration

Statement history

statementLog() returns a query builder on system.statement_info. By default it keeps only your application's statements (not MatrixOne's internal SQL), on the connection's database, from the last hour:

use Illuminate\Support\Facades\DB;

$log = DB::connection('matrixone')->statementLog();

// The 20 slowest statements of the last hour, of any type
$log->slowerThan(500)->slowest()->summary()->limit(20)->get();

// Failed statements of the last day, newest first
DB::connection('matrixone')->statementLog()->since('1d')->failed()->summary()->latest('request_at')->get();

// Slow UPDATEs on every database since a given time
DB::connection('matrixone')->statementLog()
    ->allDatabases()
    ->since(now()->subMinutes(30))
    ->ofType('Update', 'Delete')
    ->slowerThan(1000)
    ->get();

// Everything, including MatrixOne's internal SQL
DB::connection('matrixone')->statementLog()->includeInternal()->count();

Method	Effect
since('15m' | '1h' | '2d' | DateTimeInterface)	Time range (replaces the default hour)
slowerThan($ms)	duration >= $ms
failed()	status = 'Failed'
ofType('Select', 'Insert', …)	By statement_type
allDatabases() / forDatabase($name)	Replace the database filter
includeInternal() / onlyApplication()	MatrixOne's internal SQL or not
slowest()	order by duration desc
summary()	Useful columns plus duration_ms

summary() selects request_at, statement_id, statement_type, status, database, user, rows_read, bytes_scan, result_count, aggr_count, err_code, error, statement and duration_ms. The table has more: transaction_id, session_id, connection_id, response_at, duration (nanoseconds), exec_plan and stats.

Good to know:

• Times are UTC. request_at is stored in UTC whatever the session time zone. since() converts a DateTimeInterface to UTC.
• Statements appear some time after they finish: usually a few seconds, sometimes more than 15.
• Short, repeated statements are merged into one row whose text starts with /* N queries */: aggr_count is N, and duration is their total.
• Always filter by time (since() does). The table grows quickly (hundreds of thousands of rows a day on a busy server), and an unfiltered scan takes seconds.
• MatrixOne's documentation creates views such as slow_query for this purpose. They only cover SELECT statements above 1 second and scan the whole table. statementLog() covers every statement type, takes any threshold and filters by time.

Slow queries from the command line

php artisan matrixone:slow-queries                       # statements >= 1 s in the last hour
php artisan matrixone:slow-queries --since=1d --min=200  # >= 200 ms in the last day
php artisan matrixone:slow-queries --type=Select --type=Update --limit=50
php artisan matrixone:slow-queries --failed --since=6h   # failed statements with their errors
php artisan matrixone:slow-queries --all-databases --internal
php artisan matrixone:slow-queries --plan=01a0d1e1-e077-7c90-b6dc-66db263ca0ea
php artisan matrixone:slow-queries --database=matrixone_search   # another connection

The output lists the statement id, which --plan takes.

Execution plans

MatrixOne keeps the executed plan, with per-operator timings, only for statements running for at least one second (longQueryTime). Faster statements have none:

$plan = DB::connection('matrixone')->getStatementPlan($statementId);   // ExecutionPlan or null

foreach ($plan?->nodes() ?? [] as $node) {
    // step, id, name (Project, Filter, Join, Table Scan…), title, time_ms, wait_ms,
    // input_rows, output_rows, scan_bytes, memory_bytes
}

$plan->plan;   // the full decoded JSON

For a statement you can run again, EXPLAIN ANALYZE returns the same information immediately:

DB::select('explain analyze '.$query->toRawSql());

Table statistics

DB::connection('matrixone')->tableStats('orders');
// [
//     'rows' => 120034,           // mo_table_rows()
//     'size' => 4194304,          // mo_table_size(), bytes
//     'columns' => 8,             // SHOW COLUMN_NUMBER
//     'values' => [               // SHOW TABLE_VALUES
//         'total' => ['min' => 0.5, 'max' => 1999.0],
//         'status' => ['min' => 'cancelled', 'max' => 'shipped'],
//         ...
//     ],
// ]

DB::connection('matrixone')->tableStats('orders', values: false);   // skip the min/max scan

rows and size come from statistics that MatrixOne refreshes asynchronously: right after a large insert they may still show the old values (0 for a new table) for about a minute. Use count() when you need an exact number.

Other statistics statements work through DB::select(): SHOW TABLE_NUMBER FROM <database>.

Server logs and metrics

Where	Content
system.log_info	Server log records (level info, warn, error, panic, fatal)
system.error_info, system.rawlog	Errors and raw log/trace records
system.sql_statement_hotspot	The most expensive recent statements (time, memory)
system_metrics.*	Stored metrics: statement counts and durations, errors, connections, CPU, memory, disk, storage usage
Status port (status-port, 7001 in the Docker setups)	Prometheus metrics when enable-metric-to-prom = true
docker logs <container>	JSON logs of each service

// Recent server errors
DB::table(DB::raw('system.log_info'))
    ->where('timestamp', '>', now('UTC')->subHour())
    ->whereIn('level', ['error', 'panic', 'fatal'])
    ->latest('timestamp')
    ->limit(50)
    ->get(['timestamp', 'level', 'message']);

A statement that crashed the server (panic runtime error) is recorded in statement_info.error with its stack trace, which is useful when reporting a MatrixOne bug.

Configuration

Statement recording is configured in the [observability] section of each CN's TOML file (for example etc/cn.toml in the S3 Docker setup). Restart the server after changing it. The following keys exist in MatrixOne 4.2.4. MatrixOne does not document their defaults, so the values below are examples:

[observability]
longQueryTime = 1.0              # seconds; plans are kept for statements at least this long
disableStmtAggregation = false   # true: one row per statement, no "/* N queries */" merging
selectAggrThreshold = "200ms"    # statements faster than this may be merged
aggregationWindow = "5s"
skipRunningStmt = true
disableTrace = false             # true: stop recording statements (statement_info stays empty)
disableMetric = false
enable-metric-to-prom = true
status-port = 7001

The server log level is [log] level = "info" (debug, info, warn, error).

On the application side, Laravel's own tools measure queries from the client: DB::whenQueryingForLongerThan(), DB::listen(), Telescope's query watcher and Pulse's slow query recorder all work unchanged with the matrixone driver.