Sysload: Management Server and Console reactions could be slow and unresponsive

Document ID : KB000115851
Last Modified Date : 26/09/2018
Show Technical Document Details
Issue:
Management Server and Console reactions could be slow and unresponsive.

This warning appears in  Management Server log file :
 
MONTH DD HH:MM:SS YYYY: sldmgts - Could not process incoming 'alert' or 'other' connection (21464) from XXX. XXX.XXX.XXX on XXX. XXX.XXX.XXX:9901, work queue is full.
 
 
Environment:
  • Management Server
  • Analyst Rich Client (Console)
Cause:
Default settings are not optimized for important number of agents and Console Clients.
Resolution:
Some Management Server settings must be modified to be able to handle important load.
 
[General] NoCnxThreads
#Size of the connection table. 2 * Maximum number of client connections. If the maximum is reached, no more client can connect to Management Server. Its value should match the maximum number of clients (consoles, alert managers) likely to connect to the MS multiplied by 2
#Integer
#Optional
#The program defaults the value to "10" (5 clients) if not set in the configuration program.
#For 60 consoles, a value of 140 will leave a safety margin for 10 more consoles)
[General] CnxQueueSize
#Buffer size for queuing queries issued by clients (for any query type)
#Integer
#Optional
#The program defaults the value to "50" if not set in the configuration file
[General] NoAlertSubscriberThreads
#Number of threads (thread pool size) who manage Alert subscribtions from clients
#Integer
#Optional
#The program defaults the value to "20" if not set in the configuration file
#Since v5.20, this thread pool is unique and common to all clients. Before there was one pool per client
 Example of settings for 1000 Monitored Objects and 90 clients connexions :

[General]
NoCnxThreads=100 
CnxQueueSize=250
NoAlertSubscriberThreads=50
 
 
Additional Information:
Information extracted from Sysload Documentation which is available here