Re: snmpd package do not have processor load in snmp tree?

[brian moore <bem@cmc.net>]

On Fri, Jan 12, 2001 at 09:40:45AM +0800, vincent@aib.net wrote:
> Hi Brian,
> 
> Yes we can also find the load average at /etc/snmpd/snmpd.conf.
> As you said, we'd like to get the CPU usage in % (1-100%) for user,
> system, idle, as we seen from the 'top' command. Is it the per-second
> CPU usage you mentioned?

You can not get those values from snmpd, as it has no way of fetching
them from the kernel.

> We'd like to use the value in MRTG report. If we can't get it from SNMP,
> perhaps we better goto to write a script to get it from somewhere else
> (proc filesystem?). :(

Again, as I said, You can -not- get that from the kernel.

The ONLY way to do it is to read /proc/stat, sleep for a while, then
read it again.  But that is only "cpu usage for the period I was
sleeping".  Are you willing to sleep for 5 minutes to find the average
cpu over the last 5 minutes?  Willing to make your snmpd do that?
And, of course, willing to rewrite your snmp client to know that
querying that value -will- take several seconds (or 5 minutes?) to
calculate and that it should not assume the UDP packet was lost and
retransmit it....  Oh, and a way to distinguish between "this query will
take a while" and "this query must have been lost"

The kernel has -no- "how busy was I in the last <interval> period".

You can't query a value that doesn't exist.  You -will- have to sleep
for whatever interval you want to collect statistics for, and that sleep
is not likely to work with a UDP based protocol like SNMP, which assumes
a timeout means a packet was lost on the network.

-- 
CueCat decoder .signature by Larry Wall:
#!/usr/bin/perl -n
printf "Serial: %s Type: %s Code: %s\n", map { tr/a-zA-Z0-9+-/ -_/; $_ = unpack
'u', chr(32 + length()*3/4) . $_; s/\0+$//; $_ ^= "C" x length; } /\.([^.]+)/g;