ja:documentation:pandorafms:technical_annexes:50_new_architecture_784

文書の過去の版を表示しています。


サーバアーキテクチャ (バージョン 784)

In the latest change in architecture, which moved from threads to processes, it became clear that there was excessive consumption and complexity in optimizing the environment due to the existence of too many server subtypes.

アーキテクチャ の最新の変更にて、スレッドからプロセスに移行しました。サーバのサブタイプの数が多すぎることにより環境の最適化に過剰な消費と複雑さが生じることが明らかになりました。

This topic discusses a model change that mskes use of the same basis, unifying fewer servers so that having more threads for multiple tasks will make better use of hardware and time resources.

このトピックでは、同じ基盤の使用を減らし、サーバを統合して複数のタスクにスレッドを増やすことで、ハードウェアと時間のリソースをより有効に活用するモデル変更について説明します。

Servers are integrated into a single application, called Pandora Server as an ensemble, which is a multithreaded application that runs different instances or specialized Pandora FMS servers at the same time. For version 783 and earlier, the following servers were available:

  • Data Server.
  • SNMP Console.
  • Syslog Server.
  • Export Server (rarely used).
  • Sync Server (few use).
  • Discovery Server.
  • WEB Server.
  • Network Server.
  • WMI Server.
  • Plugins Server.
  • Inventory Server.
  • ICMP Server.
  • SNMP Server (different from SNMP traps).
  • Event Server.
  • Log Server.
  • SIEM Server.
  • NetFlow Server (including sFlow).
  • NCM Server.
  • Policy Manager Server.
  • WUX Server.
  • MADE Server.

Given this overview, and in an environment with half of the servers started, high memory usage was detected without having actually started working.

In version 784, the different servers have been grouped into a smaller group, which means that a producer may generate different types of tasks, e.g., a ping and a WMI check, which are now processed by the same user. This user will execute one task for WMI and another for PING, essentially doing the same thing as always.

Timeouts remain individual for each type of execution.

Network Server

It takes on the tasks of the following servers:

  • Network Server.
  • WMI Server.
  • MADE Server.
  • WEB Server.

Heavy Server

It handles the tasks of the following servers:

  • Plugin Server.
  • NCM Server
  • Export Server (disabled by default).
  • Inventory Server.

Network High Performance Server

It manages the workloads of the ICMP Server and SNMP Server.

Servers depending on a main server

  • Policy Manager (it depends on the Master server of a node).
  • Migration server (it depends on the Master server of a Command Center ).
  • Autoprovision server (it depends on the Master server of a Command Center).
  • Sync server (disabled by default).

Servers enabled by default:

  • Data Server.
  • Heavy Server.
  • Network Server.
  • Discovery Server.
  • Network High Performance Server.

Servers disabled by default:

  • MADE Server.
  • SNMP Trap Server.
  • Mainframe Server.
  • NetFlow Server.
  • Syslog Server.
  • WUX Server.
  • SIEM Server.
  • SIEM Server.
  • Event Server.
  • Log Server.
  • Alert Server.

The main configuration of Pandora FMS Server is located in file pandora_server.conf, in the default path /etc/pandora.

Starting with version 784, two main new tokens (and their thread configurations) are added:

The following tokens were removed:

Token Replaced by
exportserver heavyserver
export_threads heavyserver_threads
icmpserver networkhpserver
icmp_threads networkhpserver_threads
inventoryserver heavyserver
inventory_threads heavyserver_threads
ncmserver heavyserver
ncmserver_threads heavyserver_threads
pluginserver heavyserver
plugin_threads heavyserver_threads
predictionserver networkserver
prediction_threads network_threads
snmpserver networkhpserver
snmp_threads networkhpserver_threads
webserver networkserver
web_threads network_threads
wmiserver networkserver
wmi_threads network_threads

Separate data deletion for short intervals

In version 784 there is a new category in the deletion parameters that especially affects data with interval times lower than 300 seconds. By default, this type of data cannot be stored for more than a week, and is moved to history from that point on.

New data compression

This new data compression is used when transferring data to the history database, so that regardless of the data, no more than X samples per hour are stored in the history database (e.g., from 12, which is normal, it can change to 5 by default).

Default setting: this type of compression is not used.

If used, the default value of 5 samples per hour will be used. The general configuration control does not allow values lower than 1 or higher than 12. Compaction is not performed in cases where the number of samples is lower than the established value.

Pandora FMS ドキュメント一覧に戻る

  • ja/documentation/pandorafms/technical_annexes/50_new_architecture_784.1762689763.txt.gz
  • 最終更新: 2025/11/09 12:02
  • by junichi