need tuning help

[scuq]

hi,

Total number of objects: 62219
Number of monitored nodes: 1347
Number of collectable DCIs: 1354

cpu usage is really high, it seems like it's even higher as soon as a console-client connects:

CPU: Intel(R) Pentium(R) Dual  CPU  E2200  @ 2.20GHz

disk io, and memory seems to be ok.

i'm using postgres (tried 8.4 and 9.1) on ubuntu 12.04

what's your recommendation for NumberOf*Pollers?

br
scuq

[victor]

Hi!

Can you show how a chart of CPU usage on your NetXMS server? It seems quite strange to have very high CPU usage on that number of nodes. Regarding number of pollers, I would suggest to set it as following:

NumberOfConfigurationPollers = 20
NumberOfDatabaseWriters = 4 (or more if database writer queue still big)
NumberOfDataCollectors = 50
NumberOfRoutingTablePollers = 20
NumberOfStatusPollers = 50
NumberOfTopologyPollers = 20

High CPU usage may be caused by frequent polling of large routing tables. For example, if you have a router with full BGP table, it is recommended to turn off routing table polling for it, otherwise NetXMS server will try to fetch this huge table periodically.

Best regards,
Victor

[scuq]

thx,

is it possible to turn of routing table pollig  by default, e.g.  for s specific sysobjectid or any other filter? or should i do this with a nxshell script?

br
scuq