snmpget, snmpwalk timeout, but response comes

[Artur WILCZEK]

Hi,

I can observe something strange with new UPS equipped with InsightPower
SNMP card. I'm trying monitor it from Debian Etch (kernel 2.6.8) and
net-snmp apps in version 5.2.3. For example:

[***@ns3: ~]$ snmpget -v1 -c plmik01 ups1 system.sysName.0
Timeout: No Response from ups1.

but if I turn on debug mode...

[***@ns3: ~]$ snmpget -d -v1 -c plmik01 ups1 system.sysName.0

Sending 44 bytes to UDP: [192.168.1.100]:161
0000: 30 2A 02 01 00 04 07 70 6C 6D 69 6B 30 31 A0 1C 0*.....plmik01 .
0016: 02 04 77 64 4E 5E 02 01 00 02 01 00 30 0E 30 0C ..wdN^......0.0.
0032: 06 08 2B 06 01 02 01 01 05 00 05 00 ..+.........


Received 45 bytes from UDP: [192.168.1.100]:161
0000: 30 2B 02 01 00 04 07 70 6C 6D 69 6B 30 31 A2 1D 0+.....plmik01˘.
0016: 02 01 00 02 01 00 02 01 00 30 12 30 10 06 08 2B .........0.0...+
0032: 06 01 02 01 01 05 00 04 04 55 50 53 31 .........UPS1

As You can see, response came from UPS: 45 bytes with expected string
"UPS1".

I tried to catch responses with tcpdump:
[***@ns3: ~]# tcpdump -n host ups1
22:03:02.373063 IP 10.197.84.43.38697 > 192.168.1.100.161: C=plmik01
GetRequest(28) .1.3.6.1.2.1.1.5.0
22:03:02.381429 IP 192.168.1.100.161 > 192.168.1.43.38697: C=plmik01
GetResponse(29) .1.3.6.1.2.1.1.5.0="UPS1"

As You can see, answer looks good.

I don't know why snmpget and snmpwalk always give me timeout answer.
I think it's something specific in UPS SNMP card.

What I tried:
- on the same client upgrade snmp apps package to 5.3.1 (form Debian
Sid) - timeout, but responses are coming

- on the same client download net-snmp-5.4.tar.gz
and compile it myself - timeout, but responses are coming

- on another Debian Etch server - the same problem

- on another Ubuntu station - the same problem

- I installed net-snmp-5.4.0-1.win32.exe on Windows XP - work fine

- I installed net-snmp-5.4.0-1.win32.exe on Windows 2003 Server - work
fine as well

- I prepared simple Perl script using Net::SNMP on Debian - timeout, but
I'm able to catch responses

- set option -t with big value on Debian - it doesn't help

Generally, I can retrieve answer from UPS using net-snmp apps on Win OS,
but I have a problem to do this using net-snmp apps on Linux based OS.

There is no problems with firewalls, LAN, routing and something like that.

Thanks for all Your suggestions.

best regards,
wilku

[Dave Shield]

Post by Artur WILCZEK
Sending 44 bytes to UDP: [192.168.1.100]:161
0000: 30 2A 02 01 00 04 07 70 6C 6D 69 6B 30 31 A0 1C 0*.....plmik01 .
0016: 02 04 77 64 4E 5E ....

This outgoing request has the requestID 0x77644e5e

Post by Artur WILCZEK
Received 45 bytes from UDP: [192.168.1.100]:161
0000: 30 2B 02 01 00 04 07 70 6C 6D 69 6B 30 31 A2 1D 0+.....plmik01˘.
0016: 02 01 00....

but this response has the requestID 0x0.

Since the request ID doesn't match the ID of the original request,
this response is discarded.

The agent is broken. Complain to the vendor

It's possible that it doesn't like 32-bit IDs.
Try setting the snmp.conf directive "16bitIDs true"
(note snmp.conf, *

[Artur WILCZEK]

Post by Dave Shield
This outgoing request has the requestID 0x77644e5e
but this response has the requestID 0x0.
Since the request ID doesn't match the ID of the original request,
this response is discarded.
It's possible that it doesn't like 32-bit IDs.
Try setting the snmp.conf directive "16bitIDs true"

Yes, You are right.
Directive 16bitsIDs solved the problem.
Thanks a lot.

wilku