snmp error, output: U

[joshua]

Hi,

I have some machines which I (want to) monitor thru a OpenVPN tunnel. I am not able to retrieve SNMP data using Cacti. Neither using the poller cmd.php or spine. However I can request snmp using snmp-walk and I do get results when I manually start spine on the CLI for one of the hosts.

When spine is manually started it results:
Poller[0] Host[40] DS[1332] SNMP: v3: 10.10.10.4, dsname: mem_free, oid: .1.3.6.1.4.1.2021.4.6.0, value: 7938788

When started using the poller:
Poller[0] Host[40] DS[1332] SNMP: v3: 10.10.10.4, dsname: mem_free, oid: .1.3.6.1.4.1.2021.4.6.0, output: U

Also host detection using SNMP doesn't work, it shows the host is down with 'snmp error' however the ping method shows the host is up.

I am running Cacti/Spine 0.8.7e, using SNMP v3. I tried both poller types.

Anyone?

[gandalf]

Which parameter set for SNMP V3 are you using? Please know, that php-snmp was broken for V3 for a long time. Newest versions should have been fixed according to TheWitness.
R.

[joshua]

Hi,

I use the standard proctols (MD5/DES), time-out 500s, retries 3, max OID get size 20.
Installed php-snmp version 5.1.6-24, net-snmp 5.3.2.2-7.

So far I only saw this problem with a few servers I like to monitor. The snmp agent is running at that servers and running spine or cmd.php manually returns the correct values.

regards

[gandalf]

Wow, that's weird. So it ends up in
- manually running spine is alright
- crontab executed spine is ALWAYS wrong on CERTAIN target devices
correct?
R.

[TheWitness]

Multiple NIC's on the Cacti host?

[joshua]

@Gandalf, yep thats right. manually it runs fine, as a cron it doesn't for some devices

@TheWitness, there is one NIC. Though Cacti is running in a OpenVZ container.

Regards

[joshua]

Switching to SNMP v2 solved the problem