Skip to main content
Monitoring objects are the targets that Monitors can query and diagnose. After monit-agent is installed and started, the platform automatically displays the hosts, databases, and middleware objects that the Agent can diagnose.

​
Why monit-agent is needed

Traditional observability data usually includes metrics, logs, traces, and alert events. These data types help you understand what happened, but real troubleshooting often requires checking the live environment, for example:
  • Which processes on the current machine are using high CPU or memory.
  • Whether a port, domain, or service can be reached from the target machine.
  • Whether disks, mount points, network interfaces, or connection counts are abnormal.
  • MySQL variables, connection state, key metrics, or read-only SQL query results.
  • Redis memory, hit rate, slow logs, or read-only command results.
  • PostgreSQL connection activity, lock waits, slow queries, or read-only SQL query results.
  • MongoDB replica set status, current operations, or read-only management command results.
  • Kafka broker status, consumer lag, topic details, and group details.
  • Elasticsearch cluster health, node metrics, index statistics, and shard allocation.
Collected data alone may not answer these questions. In many cases, you need to query the live environment. monit-agent is designed for this scenario. In AI-SRE scenarios, you can think of the LLM as the diagnostic brain, and monit-agent as the execution endpoint deployed in your environment. After a user asks a question in natural language, the LLM understands the question, selects suitable diagnostic tools, and explains the result. monit-agent executes controlled queries near the target host or target service and returns structured results to the system.
monit-agent is not designed to give the LLM unrestricted execution privileges. It provides a safe, controlled, and auditable execution boundary with tool allowlists, parameter validation, command limits, timeouts, output truncation, sensitive data masking, local disable switches, and audit records.

​
Technical Architecture

monit-agent connectivity architecture monit-agent connects to monitedge in the same network region over WebSocket. monitedge then connects to the SaaS Center over WebSocket. In most deployments, each network region has one monitedge cluster, and multiple monitedge instances with the same EngineName are treated as one engine cluster.

​
Supported object types

Object typeMeaningExample address in the console
HostA server where monit-agent is installed and running10.0.1.12, host-prod-01.example.com
MySQLA MySQL instance configured in agent.yamldb-prod-01.example.com:3306
RedisA Redis instance configured in agent.yamlredis-prod-01.example.com:6379
Redis SentinelA Redis Sentinel process configured in agent.yaml10.1.2.10:26379
PostgreSQLA PostgreSQL instance configured in agent.yamlpg-prod-01.example.com:5432
MongoDBA MongoDB instance (mongod) configured in agent.yaml10.1.3.10:27017
MongoDB MongosA MongoDB routing process (mongos) configured in agent.yaml10.1.3.20:27017
KafkaA Kafka cluster configured in agent.yamlprod-order-kafka (cluster_name)
ElasticsearchAn Elasticsearch cluster configured in agent.yamlAutomatically obtained from the cluster cluster_name
After monit-agent is installed and started, at least one host object appears in the console. Other object types appear only after they are configured in agent.yaml.
Use stable and recognizable values for object identifiers (target_locator), such as fixed private IP addresses or DNS names. Do not use addresses that only make sense locally, such as localhost or 127.0.0.1, as the displayed object address.

​
How objects appear in the console

Monitoring objects do not need to be created manually in the console. Install and start monit-agent on the target machine. After the Agent successfully connects to Edge, it reports the objects it can diagnose, and the console displays them automatically. If a new user enters the monitoring object page and sees an empty list, it usually means no Agent has connected successfully. After the first Agent is connected, a host object usually appears first. If you later add MySQL, Redis, PostgreSQL, and other services to the configuration file, the corresponding objects appear automatically.
1

Install and start monit-agent

Onboard the host object first, confirm that the Agent can connect to Edge, and verify that the current host appears in the console.
2

Configure database and middleware objects

Gradually add MySQL, Redis, PostgreSQL, MongoDB, Kafka, Elasticsearch, and other objects in agent.yaml.
3

Enable controlled query tools as needed

After objects appear, enable controlled query tools such as mysql.query, redis.command, postgres.query, mongodb.command, and elasticsearch.cat only when needed.

Install monit-agent

Prepare the Edge address, download the Agent, and start it in the foreground or as a system service.

Configure Monitoring Objects

Configure hosts, databases, middleware, and custom script tools.