Connecting Data Sources

Every source reaches Energy Logserver in one of two ways. Either you install software on the source and it ships data out, or the Network Probe receives or fetches the data without anything being installed on the source at all. The first is agent-based collection, the second is agentless.

This page helps you pick between them and points to the procedure for each method. It does not repeat those procedures.

How data reaches the system

Whichever method you choose, the data lands in a Network Probe pipeline, which receives it, parses it into fields, and writes it to the Data Node. The three pipeline stages and the files that hold them are described in How data flows through a probe.

What the method decides is only the first stage, the input. An agent connects to a listener the probe already runs. An agentless source either sends to a listener or is polled by the probe on a schedule.

Choosing between agent and agentless

Method

Runs on the source

Direction

Choose it when

Trade-off

Beats agent

Beat service

Source connects to probe

You control the host and need file, event log, metric, or packet data

One service to deploy and update per host

Beats managed by MasterAgent

Beat service plus MasterAgent

Source connects to probe

You run many hosts and want to push configuration and restart Beats from the GUI

Adds a second service and a certificate to manage

Endpoint Security agent

Endpoint Security agent

Source connects to manager

You need file integrity monitoring and security configuration assessment

Heavier than a Beat, separate enrollment

Syslog

Nothing

Source sends to probe

The device can only emit syslog, which covers most network and security appliances

No delivery guarantee over UDP, parsing depends on the vendor format

Windows Event Collector

Nothing

Source pushes to collector

You cannot install software on Windows hosts

Requires certificates, group policy, and a WS-Man subscription

WinRM polling

Nothing

Probe polls the source

You need Windows data from a small number of hosts without agents or subscriptions

Polling interval limits freshness, credentials stored on the probe

Flow export

Nothing

Source sends to probe

You need traffic visibility across a network rather than logs from a host

Flow records describe connections, not payloads

SPAN through Kafka

Nothing

Probe reads the mirror

You need full traffic analysis from a mirrored port

Requires switch configuration and the bundled Kafka broker

Database polling

Nothing

Probe polls the source

The application logs to a table rather than a file

Adds query load on the database, needs a read account

API or script polling

Nothing

Probe polls the source

The vendor exposes an API, as cloud and vulnerability scanners do

Credentials live on the probe, rate limits apply

File or share

Nothing

Probe reads the path

The source already writes to a share the probe can mount

Delivery depends on the share staying available

Agent-based collection

Path

Where the procedure lives

MasterAgent and Beats

Agents

Endpoint Security agent

SIEM Agents Configuration and Endpoint Security Capabilities

Collector shipped with an integration

Advanced Installation

Beats ship with the installer at version 7.17.29 for both Windows and Linux. Winlogbeat is Windows only. The installer copies the Linux packages and the Windows archives into the agent repository that the GUI serves, so you can download them from the Console rather than from the installation media.

Only a small number of integrations bundle a collector. For every other integration the wizard has no agent step, because the source sends data on its own or the probe fetches it.

Agentless collection

Syslog

The probe listens on TCP and UDP 5514, and the installer forwards the standard port 514 to it. This is the path for firewalls, routers, switches, and any appliance that emits syslog. See Input “network” for the listener and the firewall rules, and Custom Integrations for writing a parser when the vendor is not in the catalog.

Windows Event Collector

Windows hosts forward their event logs to a collector that subscribes to them over WS-Man, with nothing installed on the source. The collector listens on TCP 5986 and forwards the events to a probe pipeline. Certificates and a group policy on the source side are required. See Event Collector.

WinRM polling

The probe runs a script that queries Windows hosts over WinRM on a schedule and turns the results into events. This suits a handful of hosts where neither an agent nor a subscription is practical. The integration is installed from the catalog, see Integrations.

Flow export

Network devices export flow records to the probe: NetFlow on UDP 2055, IPFIX on 4739, and sFlow on UDP 6343. To account for traffic that the probe host itself sees, rather than traffic reported by a device, see pmacct Local Traffic Monitoring.

SPAN through Kafka

Traffic mirrored from a switch port is analysed by the bundled sensors and carried to the pipeline through Kafka. Kafka and Zookeeper are installed but masked, and must be unmasked before use, see Kafka and Zookeeper. The sensor side is covered in Suricata IDS and Zeek.

Database polling

The probe queries a table on a schedule and turns each row into an event. MySQL, MSSQL, Oracle, and PostgreSQL are covered in Input database.

API and script polling

The probe runs a command or calls an API on a schedule. This is how cloud services and vulnerability scanners are collected. For an integration that is configured without the GUI, see Manual Configuration.

Files and shares

The probe tails a local path, or a remote share it has mounted. See Input File and Getting data from share folder.

Ports and firewall

The ports the installer opens are listed in Network Communication Requirements.

Integrations bind listener ports of their own, and the installer does not open those. After enabling an integration whose source sends data to the probe, find the port it listens on and open it:

grep -r "port =>" /etc/logserver-probe/conf.d/<integration>/
firewall-cmd --zone=public --add-port=<port>/tcp --permanent
firewall-cmd --zone=public --add-port=<port>/udp --permanent
firewall-cmd --reload

Note

Two integrations cannot share a listener port. If a pipeline fails to start with an address already in use error, check whether another enabled integration binds the same port and change one of them.

Verify the source is delivering

  1. Confirm the probe is running: systemctl status logserver-probe.

  2. Confirm the pipeline is registered in /etc/logserver-probe/pipelines.yml.

  3. Confirm the listener is bound, for a source that sends data: ss -lnp | grep <port>.

  4. Open Discover in the Console and select the index the pipeline writes to.

  5. If no documents arrive, work through Custom Integrations Troubleshooting.