SNMP not valid with poller but OK with snmpwalk & realti

[plopito]

Hi,

I have some trouble with my new cacti install.

Im trying to poll a cisco switch (snmp v3 AuthPriv) which fails each time with error :

Result from SNMP not valid. Partial Result: U

Strange things are :

- snmpwalk on the oid works
- others similar switchs are ok
- realtime plugin is ok (i see values in generated graphs)
- php cmd.php <id> is ok (grab values without problem)

I was wandering about timeouts but ping is OK and stable, and realtime never failed with a 15 seconds refresh, and snmpwalk is quick.

So how is that even possible ?
When trying manually its ok but when the poller does it's not...

(i tried to delete host & graphs and add it again -> same problem)

Thanks for any answer

[TheWitness]

Are you using Spine? If so, are you using the latest version (0.8.7.e+patches)? Realtime uses cmd.php which appears to work, but spine had an issue related to AES that was not corrected until the latest release. DES works just fine.

TheWitness

[plopito]

Hi !
Thanks for your reply.

I have same error with spine & cmd.php.
Using cacti 0.8.7e & same for spine (without patchs) on a rhel 5.3 64 bits.

I use snmp v3 this way : SHA (auth) & DES (encrypt)

These switchs have something in common : they are contacted through WAN.

Maybe it's a question of timeout ?

but :

- snmpwalk work just fine, fast. (even with a "flood" of snmpwalk)
- php -q cmd.php <id>, fast & correct
- ping is ok & stable
- links are not fully used
- tested with up to 2000ms in snmp timeout without best result (just a poller runtime longer)

What does the automatic poller (cmd.php) that the command "php -q cmd.php <id>" does not ?

[kranf]

Hi all,

I have exactly the same problem (it seems)

My cacti is Version 0.8.7e

I create 2 device (Cisco switch) and their graphs and they work fine

then I add 2 other devices (same switch model and same IOS version) which are reachable through a MAN link and i add the same graphs, i can see the graphs in realtime but they display nothing in the main window.

If someone has an idea...

I use snmpv3 with SHA and DES too

Thanks

[kranf]

for informations

the errors i can see in cacti log are :

01/08/2010 03:52:51 PM - CMDPHP: Poller[0] Host[24] DS[444] SNMP: v3: XX.XX.XX.XX, dsname: traffic_in, oid: .1.3.6.1.2.1.2.2.1.10.10128, output: U
01/08/2010 03:52:51 PM - CMDPHP: Poller[0] Host[24] DS[444] WARNING: Result from SNMP not valid. Partial Result: U
01/08/2010 03:52:51 PM - CMDPHP: Poller[0] Host[15] DS[446] SNMP: v3: XX.XX.XX.XX, dsname: traffic_out, oid: .1.3.6.1.2.1.2.2.1.16.53, output: U
01/08/2010 03:52:51 PM - CMDPHP: Poller[0] Host[15] DS[446] WARNING: Result from SNMP not valid. Partial Result: U

[TheWitness]

Please apply the spine 0.8.7e patches on the main web site. Also, use something like Wireshark to capture what's going on. But first upgrade and apply the patches.

TheWitness

[kranf]

Hi

Thanks for answering me

I just find the cacti server in that status and i have to debug it but as i m not an expert...

the server runs the cmd.php poller, i try to use spine with the patch you tell me but i lose most of graphs so i revert to cmd.php

i will try to capture some traffic to view if i can see some differences between the equipments that works and others

[kranf]

Very strange

sometimes, the graphs of the main window work for few hours and then just stop

for plopito, do you have the same thing ?

sorry for the intrusion in this topic but as i was very interesting in it, i ve begin to post

tell me if you prefer me to open a new one

[TheWitness]

Make sure you are running NTP and if on a VM, make sure you are updating all the time. This is especially true if you have a higher than average I/O or load. Don't know if that's the case. It seems highly irregular that this would be timing out.

TheWitness

[kranf]

For informations and if it can help :

I was using SNMPv3 and SNMPv2 but with many equipments (netscreen checkpoint cisco) not displaying properly their graphs.

I turn back to snmpv2 on all equipments and had to recreate config on the cisco. It seems that i have 2 problems

first, problem with the engineID on SNMPv3 (if there a way to not use it in snmpv3 ...)

second, the fact that one or more equipment were not functionning properly seem to make increase the number of processes under cacti and other equipments which were functionning correctly begin to have problem.

so i turn back on snmpv2 for the moment and will reconfigure v3 step by step later

i will try to put a feedback when i ll do the v3 config

The last question i can have is : do we have to use a unique engineid when we use SNMPv3

thanks for all

[plopito]

Hi, sorry for late answer,

No pb kranf we can share this topic, same problem same topic

But graphs never work for me in main window :/

To answer your last question, im using the same damn engineid for all cisco switchs...
Snmpv3 is pain in the xxx !

Anyway, perhaps troubles comes from the use of snmpv3 with SHA&DES.

Unforntunately, i can't use wireshark or tcpdump for the moment, but i will.

kranf : have you tried to launch cmd.php with the id of a failing switch ?
for me it works when i do that (but not with automatic polling...)

[kranf]

Hi,

I m not very familiar with cacti and don't know how to launch cmd.php with the id of a failing switch but if you can tell me i will try

For now, i have 2 C4500 and 2 or 3 C2960 which work in snmpv3 (they work since the beginning), it seems my problems are linked to the use of the engineid and maybe the view in the configuration of the cisco's.

Also we always use cmd.php cause we have many troubles with spine and snmpv3 (maybe a missing patch)

have you try to test with snmpv2 ?

[plopito]

For launching manually cmd.php with a specified host :

- get the id of the host in cacti interface
- type 'php cmd.php <id> <id>'

example : if your first switch is known with id 7, you have to type :
php -q cmd.php 7 7

I have another cacti in production, which use snmpv2 to poll same equipement and there is no problem.

[peanut3122]

having same problem..

a ubuntu vm running latest everything

All was good before moving up to e, now no data in the graphs. Well only devices polling with snmpv3, the mysql templates still looked good. snmp walk seems to work.

Other were fine, but I blew it all up trying over, so far same problem still. Upon adding device you immediately get a snmp error.

*update* snmp v2 works fine for all devices

[plopito]

Ok so it seems to be snmpv3 with cacti...

peanut3122 > do you use sha+des for snmpv3 ? have you tried to launch cmd.php manually like in my last post ?