auto-discovery wizard never finishes

[hhlodge]

Hi,

I have been trying for days to get an auto-discovery run on even one Linux host. The Auto-Discovery Jobs window just has the spinning Status wheel going for hours until I give up and kill it. A tcpdump for the host shows the same OIDs and values over and over again. Any clues what I can do to try and debug what's going on with the wizard? Below is one record I've seen over a dozen times in the past hour. This is XI 2012R2.8c on a CentOS 6.5 64-bit VM on ESXi 5.1. Thanks.

Code: Select all

16:06:10.326098 IP host.domain.snmp > nagioshost.domain.44936:  C=public GetResponse(560)  25.2.3.1.3.1="Physical memory" 25.2.3.1.3.3="Virtual memory" 25.2.3.1.3.6="Memory buffers" 25.2.3.1.3.7="Cached memory" 25.2.3.1.3.10="Swap space" 25.2.3.1.3.31="/" 25.2.3.1.3.35="/home" 25.2.3.1.3.36="/u01" 25.2.3.1.3.37="/usr" 25.2.3.1.3.38="/tmp" 25.2.3.1.3.39="/var" 25.2.3.1.3.40="/boot" 25.2.3.1.4.1=1024 25.2.3.1.4.3=1024 25.2.3.1.4.6=1024 25.2.3.1.4.7=1024 25.2.3.1.4.10=1024 25.2.3.1.4.31=4096 25.2.3.1.4.35=4096 25.2.3.1.4.36=4096 25.2.3.1.4.37=4096 25.2.3.1.4.38=4096 25.2.3.1.4.39=4096 25.2.3.1.4.40=1024 25.2.3.1.5.1=16221784

[tmcdonald]

We just released 2.9 which addresses this issue.

Code: Select all

cd /tmp
wget http://assets.nagios.com/downloads/nagiosxi/2012/xi-2012r2.9.tar.gz
tar xzf xi-2012r2.9.tar.gz
cd nagiosxi
./upgrade

[hhlodge]

Sure did - it finished in about 10 seconds. I see all the services it discovered. I don't see any file systems or other entiities. Maybe it's not supposed to capture those, even though I see the data being returned to the XI host.

[tmcdonald]

Auto-discovery just runs nmap on the backend and does not (to my knowledge) have access to internal metrics like filesystem usage or CPU load.

[hhlodge]

My bad, I was already monitoring that particular host so I was seeing normal SNMP polls from already-in-place Nagios checks. Beyond that, if the Linux SNMP wizard can query for file systems, processes, etc., surely the auto-discovery wizard could be enhanced to do so. It would make it much more useful. As it is now, just seeing open ports doesn't do much for me from a monitoring standpoint.I'd have thought it to be smarter such that it would determine the OS and if Linux, then attempt a SNMP walk and if Windows, WMI queries if SNMP doesn't answer. Both could use some default credentials set somewhere, like having them in resources.cfg. Just some thoughts.

[tmcdonald]

Great suggestions regarding SNMP/WMI. I opened a feature request for you here: http://tracker.nagios.com/view.php?id=502

We're all pretty busy getting XI 2014 ready to launch, so I can't guarantee when it will be reviewed but our devs do have new issues in their queue.