Spine 0.8.7c doing unrequested tcp ping on down hosts

[ciranor]

I've just upgraded both cacti and spine to 0.8.7c, and immediately hit a problem where 2/3 of hosts were marked as down for no apparent reason given they'd all been up a few minutes earlier under cacti 0.8.7b/spine 0.8.7a.

Running spine in verbosity=5 mode against 2 sets of 3 hosts which were either considered all up, or all down, this was present in the case where the boxes were all considered to be down:

Host[335] PING Result: TCP: Cannot connect to host
Host[335] SNMP Result: Host responded to SNMP
Host[336] PING Result: TCP: Cannot connect to host
Host[336] SNMP Result: Host responded to SNMP
Host[337] PING Result: TCP: Cannot connect to host
Host[337] SNMP Result: Host responded to SNMP

but in the case of the boxes that were all considered to be up, those TCP PING lines were not present.

The "Downed Host Detection" setting in the poller config was checked and set to "SNMP" only, but I tried changing it to other poller types without any apparent effect.

I've downgraded spine back to 0.8.7a and it is happily collecting data once more. Was there something I missed out on in the upgrading of spine, or is there a bug in the downed host detection?

[TheWitness]

What was the host config set to? What version of Cacti are you running?

TheWitness

[ciranor]

Whoops, that makes me feel stupid. I'd never registered that there was a separate per host "downed host detection" setting.

Sorry for the noise.