Is netxms capable of handling 2000+ monitord nodes ? I have a dual 4 core server with 20gigs of ram and netxms seems to be choking I constantly get down events of devices that are not down .. it seems its not handling this very well What would be the suggested variables for such a setup ? Im using percona server for the db which is supposed to be a more optimized mysql replacment....
Thanks
forgot to mention it takes about two minutes or so for the server to even start and after that I get a hole bunch of events complaining about nodes being down in the gui
What version are you running and what OS?
netxms 2.0 ubuntu 12.10
have you tried increasing the number of pollers ?
you can do it in the "server configuration"
some time ago I had a similar problem after upgrading libmysql.
I solved this by increasing these parameters:
* ConnectionPoolBaseSize to 20
* ConnectionPoolMaxSize to 500
With the growth of my network I increased these parameters to:
* NumberOfConfigurationPollers, NumberOfStatusPollers, NumberOfDataCollectors
* NumberOfDatabaseWrites to 4
Best regards
Georg
OK I'll give that a shot how many nodes do you monitor... and howe often do you poll for status
Still nothing... at this point I cant really trust what netxms is telling me every time a node shows down I can still ping can communicate with
I really suggest you contact Raden and get official support for the issue.
They will be able to solve it fast.
ok so far so good I pretty much wiped the original server which had Ubuntu 12.04 and installed to latest version and installled percona server again and after doing a little tuning it seems to be doing great Im currently at 900 nodes and it hasnt skipped a beat I will slowly add more nodes at a time and see when I start to see issues but I think it was just a lot of bad mojo that needed to be wiped out .. Ive used this as a test server for a lot of things so I think it was time.. Im just curious as to what is the max number of nodes anyone has used netxms for in a production environment and what type of server etc... was used
I currently have one deploy at 1k nodes and 2 other at 300 nodes.
You will find people here on the forum with 3k+ node deploys.
Can someone elaborate on the following error ? Looking at the server load itself its consistently at about 0.22
What is the POLLERS thread pool size?
General rule is load average should never be higher than current pool size.
Usage of the thread pool should also never reach 100 (its percentage), if it does consider increasing PollerThreadPoolMaxSize server config variable.
Base = 120
previous max = 240 restarted server
previous max =1200 restarted server
current Max = 2500 restarted server
my starting max was 240 so i gradually increased it to its current 2500 witch is well over 1/10 the amount of nodes. And its seems to always give that same error no matter how much i increase its but the usage did go from 99 down to 43
Server thread pool POLLERS: usage = 42
Server thread pool POLLERS: load average (1 minute) = 32.141113
Server thread pool POLLERS: current size = 1066
Server thread pool POLLERS: current load = 6
Hi,
load average 32 for a pool means that you have about 32 tasks running in parallel on this thread pool. You can check history for that parameter and set pool max size at a bit higher value then max for load average and min size to medium value for load average.
Best regards,
Victor
Hello gmonk63,
are you polling all your nodes directly from the NetXMS Server or are you also using proxy agents? I advise you to use nodes proxy agents according to the subnets you are monitoring. There is a parameter in nxagentd.conf called MaxSessions. The default value is:
MaxSessions = 32
The highest value of max. sessions can be set till 1024. So try to set:
MaxSessions = 1024
But after that. As I already wrote. It's better to set up nodes as proxy agents.
Greets,
Andreas