UNKOWN hostname when restarting gmetad

[patricioe]

UNKOWN hostname when restarting gmetad

[scottbessler]

I work with patricioe, we see UNKNOWN in nagios because ganglia (not sure if its gmond on the box that is being monitored, or something on the box that receives the metrics) is not honoring the host name spoofing setting on the box that is being monitored (and thus doesnt send the right name to nagios)

The steps to reproduce seem to be:

1. Start gmond gmetad on monitoring-server
2. Start gmond on watched-instance (everything seems to work at this point)
3. Restart gmond/gemtad on monitoring-server
   and notice that the boxes are being reported as actual hostname instead of spoofed name.

[jbuchbinder]

Are you spoofing all of the traffic using the newer gmond.conf spoof settings, or are you feeding spoofed metrics via gmetric and its -S parameter?

[scottbessler]

gmond.conf

[scottbessler]

we push some metrics with gmond (load, etc.) but we also do push some via gmetric, expecting it to use the gmond.conf (which it usually does except in the described scenario)

/usr/bin/gmetric --dmax=3600  --group=process --name=$1_status --type=uint8 --value=$retval

[patricioe]

we removed override_ip info and it survived after a couple of gmetad restarts. Coincidence or might it be related to the issue ?

[jbuchbinder]

There was a recent patch incorporated into the master branch which fixed an override_ip issue; could you retest?