makinacorpus/monitoring-bundle

Simple monitoring tooling and Symfony bundle

1.0.0-alpha7 2023-09-05 12:33 UTC

This package is auto-updated.

Last update: 2024-05-08 15:02:08 UTC


README

This API provides simple and easy to use interfaces for building supervision probes that you can run to check sanity of your application.

It provides a Symfony >= 4.4 compatible bundle for registering your custom probes into a Symfony application.

Features:

  • Build probes that restitute an output similar to commonly used supervision tools, such as Nagios or Centreon.
  • Build more advanced information collectors for building advanced status reports.
  • Provide an easy to use probe and info collectors registry.
  • Provide a few console commands to run probes or build status reports.
  • When used via the Symfony bundle, provide easy probes registration and HTTP endpoints for querying probes: easy to use with supervision tools.

Installation

Install it using composer:

composer req makinacorpus/monitoring-bundle

Symfony bundle

Installation

Then register the bundle in your app/bundles.php:

<?php

return [
    // Other bundles...
    MakinaCorpus\Monitoring\Bridge\Symfony\MonitoringBundle::class => ['all' => true],
];

Configuration

Optionnaly copy the src/Bridge/Symfony/Resources/config/packages/monitoring.yaml file into your app config/packages/ folder.

Register HTTP status endpoint (recommended)

Status endpoint will always return plain text responses, Nagios parser compatible, which means that almost every open source supervision tool will understand those status reports.

A custom route loader will generate one route per probe. In order to register those routes, add into your config/routes.yaml the following code:

monitoring_endpoint:
    resource: "@MonitoringBundle/Resources/config/routes/endpoint.yaml"

For those routes to respond, you need to generate an access token for users:

bin/console monitoring:generate-token

Follow the instructions on screen, it will display the new token and new probes URL: you can copy/paste the newly generated token into your environment variables:

MONITORING_TOKEN=your-generated-token

Note that you can run the command as many times as you wish in order to be able to copy/paste probes URLs, the command will never modify your application configuration.

If your site is protected by a firewall, you may add the following into config/packages/security.yaml:

security:
    firewalls:
        # Monitoring (access is token protected within controller)
        monitoring:
            pattern: ^/monitoring/status
            security: false

Register admin report screen (optional)

A basic HTML report controller and template is provided, you may add it into your config/routes.yaml configuration file:

monitoring_admin:
    resource: "@MonitoringBundle/Resources/config/routes/admin.yaml"
    prefix: /admin

Please note that it is not security checked, you must manually configure your firewall to protect it.

Building your own probes and reports

Build a simple probe

Implement the \MakinaCorpus\Monitoring\Probe interface.

declare(strict_types=1);

namespace App\Monitoring;

use MakinaCorpus\Monitoring\Probe;
use MakinaCorpus\Monitoring\ProbeStatus;

/**
 * Collects number of items to process in queue, raise error when it is too much.
 */
final class QueueSizeProbe implements Probe
{
    /** @var ?int */
    private $warningThreshold;

    /** @var ?int */
    private $criticalThreshold;

    /** @var \My\Favourite\Database\Client */
    private $database;

    public function __construct(
        \My\Favourite\Database\Client $database,
        ?int $warningThreshold = null,
        ?int $criticalThreshold = null
    ) {
        $this->database = $database;
        $this->warningThreshold = $warningThreshold;
        $this->criticalThreshold = $criticalThreshold;
    }

    /**
     * {@inheritdoc}
     */
    public function getName(): string
    {
        // Internal name.
        return 'queue_size';
    }

    /**
     * {@inheritdoc}
     */
    public function getTitle(): string
    {
        // Human readable name, for reports.
        return "Queue size";
    }

    /**
     * {@inheritdoc}
     */
    public function getStatus(): ProbeStatus
    {
        $queueSize = $this->database->query('SELECT COUNT(*) FROM "my_queue_table" WHERE "is_consumed" is false')->fetch();

        // For those who know Nagios or the like, just set a very short status
        // message intended for display purpose in larger reports.
        $message = \sprintf("my queue size: %d items", $queueSize);

        if ($queueSize >= $this->criticalThreshold) {
            return ProbeStatus::critical([$message, "queue is failing !"]);
        }
        if ($queueSize >= $this->warningThreshold) {
            return ProbeStatus::warning($message);
        }
        return ProbeStatus::ok($message);
    }
}

Build a simple report generator

Implement the \MakinaCorpus\Monitoring\InfoCollector interface:

declare(strict_types=1);

namespace App\Monitoring;

use MakinaCorpus\Monitoring\InfoCollector;
use MakinaCorpus\Monitoring\Output\CollectionBuilder;

/**
 * Collects information about all database tables.
 */
final class DataInfoCollector implements InfoCollector
{
    /** @var \My\Favourite\Database\Client */
    private $database;

    public function __construct(\My\Favourite\Database\Client $database)
    {
        $this->database = $database;
    }

    /**
     * {@inheritdoc}
     */
    public function getName(): string
    {
        // Internal name.
        return 'data';
    }

    /**
     * {@inheritdoc}
     */
    public function getTags(): iterable
    {
        // Tags will help build specific reports.
        return ['database', 'data', 'volume'];
    }

    /**
     * {@inheritdoc}
     */
    public function getTitle(): string
    {
        // Human readable name, for reports.
        return "Data information";
    }

    /**
     * {@inheritdoc}
     */
    public function info(CollectionBuilder $builder): void
    {
        // Yes, this should work with pgsql.
        $rows = $this->database->query(<<<SQL
SELECT 
    pg_size_pretty(pg_total_relation_size(relid))
        AS "total size",
    pg_size_pretty(pg_table_size(relid))
        AS "table size",
    pg_size_pretty(pg_indexes_size(relid))
        AS "index size",
    concat(schemaname, '.', relname),
    concat('seq_scan: ',seq_scan, E'\\nseq_tup_read: ', seq_tup_read, E'\\nidx_scan: ', idx_scan, E'\\nidx_tup_fetch: ', idx_tup_fetch)
        AS "reads",
    concat('insert: ',n_tup_ins, E'\\nupdate: ', n_tup_upd, E'\\nhot_update: ', n_tup_hot_upd, E'\\ndelete: ', n_tup_del)
        AS "writes",
    concat('live: ',n_live_tup, E'\\ndead: ', n_dead_tup, E'\\nmod_since_analyze: ', n_mod_since_analyze)
        AS "state",
    concat('last_vacuum: ',last_vacuum, 'auto: ', last_autovacuum, E'\\nlast_analyze: ', last_analyze, ' auto: ', last_autoanalyze, E'\\ncpt_vacuum: ', vacuum_count, ' auto: ', autovacuum_count, E'\\ncpt_analyze: ', analyze_count, ' auto: ', autoanalyze_count)
        AS "vacuum"
    FROM pg_stat_user_tables
    ORDER BY pg_total_relation_size(relid) DESC
SQL
        );

        $table = $builder->addTable()->setHeaders([
            'table', 'total_size', // ...
        ]);

        foreach ($rows as $row) {
            $table->addRow([
                $row['relname'],
                $row['total size'],
                // ...
            ]);
        }
    }
}

Combining both

Just implement both interfaces, they are compatible and won't conflict.

Notes about reports and tags

Each InfoCollector implementation has a getTags(): iterable method, each tag is a tag name string. Beware that you probably want to group associated reports altogether for building your UI.

Registering them into Symfony

For any probe or info collector class, register them into your container and add the monitoring_plugin tag:

services:

    # Considering that auto-wiring is enabled.
    _defaults:
        autowire: true

    App\Monitoring\:
        resource: '../src/Monitoring'
        tags: ['monitoring_plugin']

Cron script

If you don't have a supervision tool to parse status endpoint, you can use a simple check command run by a cron to perform status check.

Configuring your cron

If you don't have any external software to read your probes, you can plug-in the default (very naive) version of monitoring daemon into your crontab:

# Every 10 minutes: sanity check.
# In real life, this is the job of your sysadmin to choose recurrence.
*/10 * * * * /path/to/your/symfony/app/bin/console monitoring:check

Pluging reactions to errors

Per default, the check command won't do anything, you must manually write handlers for reacting upon broken probes.

<?php

declare(strict_types=1);

namespace App\EventSubscriber;

use MakinaCorpus\Monitoring\Event\ProbeResultEvent;
use Symfony\Component\EventDispatcher\EventSubscriberInterface;

/**
 * @codeCoverageIgnore
 */
final class MonitoringEventSubscriber implements EventSubscriberInterface
{
    /**
     * {@inheritdo}
     */
    public static function getSubscribedEvents()
    {
        return [
            ProbeResultEvent::class => 'onProbeResult',
        ];
    }

    public function onProbeResult(ProbeResultEvent $event)
    {
        if ($event->isCritical()) {
            // Do something.
        } else if ($event->isMalfunctioning()) {
            // Do something else.
        }
    }
}

Fetching nagios-compatible result using the command line

/path/to/your/symfony/app/bin/console monitoring:status

This will output a Nagios parser compatible text.