Problem with ucd/net SNMP host Memory Usage Graph

[anubis]

Hello, i'm hoping someone can help with this.

I have 7 HPUX servers running 11i and net-snmp-5.2.1.2

6 out of the 7 work perfectly for ucd/net cpu load and memory

one of the servers, the memory usage graph is always empty however the cpu useage and load average work fine. and its only the one server exhibiting this, the other 6 work fine.

all servers are running the same OS and net-snmp, and perl 5.8.0

whats the best way to debug this?

thanks in advance!

Aaron

[anubis]

oh
and the cacti version i am running is cacti-0.8.6g-1.2.el4.rf

[gandalf]

one of the servers, the memory usage graph is always empty

Look at the appropriate data template. You'll find the OID for memory usage there. Try an snmpwalk for that OID from cli. Suppose it will get no response. Then, please have a look at the snmp config of your machines. As far as I know, memory OIDs do not pertain to standard MIBs. Perhaps the specific MIBs are missing on that machine...
Reinhard

[anubis]

one of the servers, the memory usage graph is always empty

Look at the appropriate data template. You'll find the OID for memory usage there. Try an snmpwalk for that OID from cli. Suppose it will get no response. Then, please have a look at the snmp config of your machines. As far as I know, memory OIDs do not pertain to standard MIBs. Perhaps the specific MIBs are missing on that machine...
Reinhard

thanks for the reply - i will check it out - the thing that confuses me is all 7 servers are configured the same, and the other 6 work fine.

I will check it out and get back to you.

thanks again!

[anubis]

Look at the appropriate data template. You'll find the OID for memory usage there. Try an snmpwalk for that OID from cli.

the oid is .1.3.6.1.4.1.2021.4.6.0

so what would be the syntax of the snmpget command if i want to specify the udp port number?

i guess i should mention that this 1 server that has this issue has net-snmp listening on udp 163 rather than udp 161 (because this server has HP Network Node Mgr running on it and i do not want net-snmp to interfere with it)

i do have port 163 specified in the device configuration and it does successfully retrieve all the other OIDs - just not the memory one.

do you know how to specify the udp port on the snmpget command?

[anubis]

the snmpget does return an integer:

[root@xlin101 ~]# snmpget -v 1 -c [snip] xnnm000:163 .1.3.6.1.4.1.2021.4.6.0

UCD-SNMP-MIB::memAvailReal.0 = INTEGER: 11186076

so now i am confused

and yes i have deleted and rebuilt the graphs

if it is kicking back an integer, why isn't it populating the graph?

[gandalf]

Ok, fine. So please lets have a look at the log. Please switch to DEBUG mode (under "Settings") and run at least one polling cycle. Have a klook at the <path_cacti>/log/cacti.log. You'll find your OID there. Around, you'll find what data was returned and you should find an appropriate "rrdtool update" statement. If this is there, ok. If not, you will perhaps have to "Clear Poller Cache" (see "System Utilities").
If the update is shown but is not working, this may be due to wrong permissions on that very rrd (often happens to me if I run the poller under root for debugging). You may run the poller cmd (see crontab) after su-ing to <cactiuser> and have a look at the output ...
Reinhard