Version

Argent Guardian Ultra 3.1A-1410-A And Later

Date

Monday, 17 Nov 2014

Summary

When Argent monitors Linux/UNIX servers through SSH, the PLINK processes can consume a lot resource on Main Engine

SSH Relay Agents can be used to move the load off the Main Engine

In earlier versions, each monitored Linux/UNIX server can be only associated with one SSH Relay Agent

This sometimes did not scale well nor provided failover when the SSH Relay Agent is down

The SSH Relay Agents feature has been enhanced in Argent AT 3.1A-1410-A

Customers can specify multiple SSH Relay Agents in the Supervising Engine screen

The names of multiple SSH Relay Agents are separated by commas

Customers can also select option ‘Randomly ‘or ‘Sequentially‘

Randomly Agents are randomly selected so that load is always distributed evenly

Sequentially The first available agent is selected from the list

When the selected agent is not available, the next available agent is selected from the list

At the Node level, customers can specify property ‘Relay Agents‘ to one of following values:

{default}

Use the default setting specified in Supervising Engine screen

In other words, the node will be monitored using the default SSH Relay Agents specified in Supervising Engine

See the highlighted field ‘Use Default SSH Relay Agents‘ in previous screenshot

{none} or empty

SSH Relay Agent is not used

The node will be monitored using the default SSH method by spawning a PLINK process at Argent AT engine

It is true even there are some default SSH Relay Agents specified in Supervising Engine screen

Explicit IP address for SSH Relay Agent

Explicit agent should be used

This overrides the default SSH Relay Agents specified in Supervising Engine screen

The node will be always monitored using the SSH Relay Agent explicitly specified here

Technical Background

When Argent Monitoring Engine monitors Linux/UNIX servers by PLINK.exe, it has to spawn a PLINK.exe process for each monitoring task

In Windows 2008 and later, each PLINK process has an accompanying conhost.exe process

This is true no matter if pooled Monitoring Engine process is used or not

When Argent monitors many Linux/UNIX servers, the sheer amount of PLINK processes can put a heavy load on the system

Due to the implementation of open SSH, Linux/UNIX server generally scales a lot better than PLINK from a Windows server

As a result, SSH Relay Agent can handle more monitoring tasks and also move the load out of Main Engine

Resolution

Upgrade to Argent Advanced Technology 3.1A-1410-A or later