A SpectroSERVER in a large distributed cluster fails to discover device models.
The Discovery Configuration appears to make it to model creation but before it moves to the next stage, before new models are found, the SpectroSERVER crashes.
Restarting the SS works fine. It restarts but repeatedly crashes during run of Discovery Configuration.
Same discovery on other SS's in the environment work fine.
The VNM.OUT shows the following error.
Oct 31 12:13:45 ERROR TRACE at IcmpServer.cc(831): Unable to get address information for ICMP listener port of type - -2(Name or service not known)
We also notice there are no raw sockets opened that would normally be required for ICMP. That's seen from "netstat -anop | grep raw" returning nothing. Working SS systems will raw socket ports open and ready for use.
All supported Network Observability DX NetOps Spectrum SpectroSERVER releases
A $SPECROOT/SS/.vnmrc variable for icmp is incorrectly configured.
Ensure the icmp_listen_ip variable value in the .vnmrc value is properly set.
There are two options:
Any other value, anything outside of a blank or a valid IP, will cause the VNM.OUT errors and SpectroSERVER failure during discovery run attempts.