snmp issues

[navtek007]

just trying to queiry our pix via snmp and it keeps on failing. the error logs look like the following:

4924:20071207:135339 Error doing snmp_open()
4924:20071207:135339 Parameter [ifInOctets.2] is not supported by agent on host [Firewall] Old status [0]

I have snmp running on other routers fine with no problems. I can do a snmp get from the console with no issues as well.

any other ideas?

[Tenzer]

Are you sure that the Zabbix server is compiled with SNMP support?

[martin.marcher]

4924:20071207:135339 Parameter [ifInOctets.2] is not supported by agent on host [Firewall] Old status [0]

• try using the real OID.
• double check the port (I know but I just had that)
• http://www.zabbix.com/wiki/doku.php?id=contrib:snmp
  • oh and please add how to configure your device and zabbix if you can

[claytronic]

I am also having trouble with SNMP checks of a PIX. I've upgraded to 1.4.3 (very pleased with changes and fixes) on both master and remote nodes. The master is able to perform SNMP checks just fine. The remote reports "Error doing snmp_open()". I've confirmed that the remote can successfully snmpwalk/get from the command line.

Both zabbix servers are running Ubuntu 7.10 with the same packages and compiled with:

./configure --enable-server --enable-agent --with-mysql --with-net-snmp --with-libcurl

Any idea's?

[claytronic]

[Update]

Forgot to switch from DNS to IP as the "Connect to" method. SNMP checks work so much better when there's a valid host to query.

[css1971]

Problem with SNMP dropping monitoring items.

On the item configuration screen, zabbix reports:

Processor load: CPU 0 processor-load-cpu0 100 90 365 SNMPv2 agent Not supported CPU Items Error doing snmp_open()

In the log it reports:

14245:20080507:180941 Parameter [processor-load-cpu0] is not supported by agent on host [host.name.here] Old status [0]


and the config of the processor-load-cpu0 item is as follows:

Processor load: CPU 0
SNMPv2 agent
commiunitystring
.1.3.6.1.2.1.25.3.3.1.2.768
161
processor-load-cpu0
Numeric (integer 64bit)
% Busy

If I query the host directly using snmpget, it works fine.

snmpget -v 2c -c communitytring host.name.here .1.3.6.1.2.1.25.3.3.1.2.768

HOST-RESOURCES-MIB::hrProcessorLoad.768 = INTEGER: 6

Now. Zabbix (1.4.5) is configured with SNMP, the same template works with a bunch of other hosts. The difference with this particular machine is that it is a development machine and tends to be up and down like a yoyo.

I think therefore that zabbix is incorrectly marking items as "Not Supported" and stops monitoring when it is unable to perform the SNMP query, even though the host may simply be down.

This is rather a concern because a significant part of using a monitoring system is to tell you when something is unavailable.

I haven't yet started to trace through the code but have increased the debug level.

[Robert Wagnon]

Detect down with .nodata

How about adding a Trigger with keyname.nodata(300)? Then you know you didn't get information from your Item.

[css1971]

How about adding a Trigger with keyname.nodata(300)? Then you know you didn't get information from your Item.

Use Zabbix to tell me when Zabbix has stopped working?

Sorry, doesn't make sense, if it has stopped working, how can I rely on it to tell me it has stopped working?

[css1971]

Refresh Unsupported

Tracing the poller code for snmp, it looks like it sets an item to unsupported if the host is up but the snmp query fails. There is some code which uses CONFIG_REFRESH_UNSUPPORTED to re-try the check at a later time. However that config entry in the config table simply didn't exist in my case, though I had run the mysql patch for the update. I don't think the config entry existed in the the 1.1 system.

Anyway. I've added a line to the config table and set the refresh unsupported to 14400 seconds so I'll see if unsupported problem goes away after I reactivate all the polled items.

mysql> select * from config;
+----------+---------------+---------------+---------------------+------------------+----------------+
| configid | alert_history | event_history | refresh_unsupported | work_period | alert_usrgrpid |
+----------+---------------+---------------+---------------------+------------------+----------------+
| 1 | 365 | 365 | 14400 | 1-5,00:00-24:00; | 8 |
+----------+---------------+---------------+---------------------+------------------+----------------+

[css1971]

Yup, seems to work

the refresh unsupported option causes zabbix to restart monitoring of the "unsupported" items. I think i'll drop the time to one hour though.

Happy camper.