Downed Host Detection Catch22?

[echochargen]

Hello all. Thanks for this great piece of software.

I'm attempting to monitor my 2 squid caches with snmp via port 3401 (default squid snmp port). I've installed the squid templates from Rolf's excellent post (http://forums.cacti.net/viewtopic.php?t=3158).

I have my squid device created to use port 3401. The top left corner shows an "snmp error" message, but the verbose query via port 3401 is successful. When the poller runs, it marks the device as "down", and doesn't continue to poll the snmp values. If I follow the advice from this thread (http://forums.cacti.net/viewtopic.php?t=4799) and remove the community string, then the community string isn't sent during the polling cycle (I see a value of ' ' in the squid debug log).
Is it possible to accomplish this? Is there some way to have this device ignore the "downed host detection" routine?

Thanks for any input you can provide.

Best Regards,
Jeremy

[TheWitness]

Use Ping only host availability and I think you will be all set.

TheWitness

[Guest]

I think his problem is that he wants to monitor via SNMP, but the device doesn't give the SysDescr on port 3401.

This is the same problem I'm currently having with our Nokia/Check Point firewalls. Nokia makes the hardware, and you load the Check Point software onto it. The Nokia appliance listens on the standard port (161) and is what you need to poll to get interface statistics. The Nokia appliance populates the SysDescr field, so everything works fine.

The Check Point firewall software also runs SNMP, but it runs on port 262. This SNMP instance contains nothing except firewall related statistics such as the current entries in the state table, etc. Since it does not contain data for the SysDescr field, cacti thinks the host is down and will not poll the other configured OIDs.

I looked for a solution a few weeks ago, and I think the next vesion of Cacti will let you disable the down host detection on a per-device basis. I don't believe there is a way around this in the current version.

[cacti_fod]

Yes, I'm running into this annoying limitation as well. I'm graphing interface data via SNMP, but I'm also graphing RTT/Packetloss and when the host goes down I want that 100% packetloss to be graphed.

Has anyone hacked anything together to disable the host down stuff?

If checks are going to down a host because of availability issues, then there should be a lot more logic to the mechanism so that you can tune it to your environment ( like turn it all off, turn off just SNMP checks until a check is good.... but then you start getting more into monitoring and not graphing ).

[TheWitness]

You can modify the cmd.php to check another non-blank OID instead of ifDescr. But select one that is general enough to work for all hosts. I am not too sure which one it would be. Then should be Ok. This issue will be corrected in 0.8.7.

TheWitness

[alamp]

Hi!

I ran into the same problem, changed the poller "downed host detection" to UDP ping and cacti started to draw some nice graphs.

Anyway, my pix firewalls which dont answer to UDP pings, went down 8(

So i made a very little change to script lib/ping.php:

on line 445, i deleted the part "&& ($ping_result == true)" to make the line look like this

if ( ($avail_method == AVAIL_SNMP) || ($avail_method == AVAIL_SNMP_AND_PING) ) { /* make OR instead of AND */

I've also changed the downed host detection to "SNMP and ping".

This checks now with UDP ping AND snmp "ping" and if one of this checks is succesfull, the host is up.

This solved my problem till the new shining version will appear

Greetings

[gandalf]

Along with the original script you may find a description on how to use snmp proxy function. This will allow to send the request and stuff on port 161, while the procy switches it to 3401. We're using this since ages with success. It requires net-snmp on the target host
Reinhard