snmpget output parser

[user]

Hello.
I've some managed switch which have Counter64 type of IF-MIB::ifInOctets.* instead of Counter32 and snmpget work like this:

Code: Select all

# snmpget -v 2c -c public 192.168.200.2 .1.3.6.1.2.1.2.2.1.10.5
IF-MIB::ifInOctets.5 = Wrong Type (should be Counter32): Counter64: 5600637353944

It appends "Wrong Type (should be Counter32): ", which I think is becoming a parse error in Cacti and it shows "U" during updates in logs.
Can You take this case into account in the future versions of Cacti?
PS. My version of Cacti is 0.8.6f (last at this moment), if this matters.
PPS. Switch have nothing at IF-MIB::ifHCInOctets.*, i.e. where really Counter64 should be, as I know.

[user]

Here is a patch, attached. As I understand, function strip_quotes only cut nonnumeric data at the end of the output. This adds cutting beginning stuff.

[user]

Hmm... things are more complex, that I thought. My patch brokes data inputs returning several values.
So, I can only suggest this workarond-like patch. I hope it will not brake something else.

[rony]

When I looked at you initial patch, I thought to myself, "this is going to break other stuff". And it did, you beat me to posting..

Your second patch is a much better approach. And btw, I have this problem with some IBM agents, but I didn't think of parsing it out, I called IBM and told them to fix the MIB.

[paulbeard206]

It might trim off all that extraneous text: I don't have a 64 bit device to test it but have used that combination before.

[TheWitness]

Fixing this in the poller will cause a 5-10% performance penalty during polling. I would rather see the vendor fix it.

TheWitness

[rony]

Ok, you said 5% to 10%, that is a problem, 1% to 2%, would not be such a problem.

Thanks,

[TheWitness]

I could do it and then you could test? Would you like cmd.php or cactid?

TheWitness

[jrichardson]

What is the fix for this?

I've got a cisco 6509 and when I try to poll the 64bit counters, I also get the "Wrong Type (should be Counter32):" text.

I'm running 0.8.6c.

[jrichardson]

What is the state of this patch?

I've got 0.8.6f-r1 installed on a gentoo box and it's got:

Code: Select all

function strip_quotes($result) {
        /* first strip all single and double quotes from the string */
        $result = strtr($result,"'","");
        $result = strtr($result,'"','');

        /* clean off ugly non-numeric data */
        if ((!is_numeric($result)) && ($result != "U")) {
                $len = strlen($result);
                for($a=$len-1; $a>=0; $a--){
                        $p = ord($result[$a]);
                        if (($p > 47) && ($p < 58)) {
                                $result = substr($result,0,$a+1);
                                break;
                        }
                }
        }

        return($result);
}

but on a 0.8.6c box it's simply:

Code: Select all

function strip_quotes($result) {
        /* first strip all single and double quotes from the string */
        $result = strtr($result,"'","");
        $result = strtr($result,'"','');

        return($result);
}

[en3my]

'lo there...

Same problem here with some rubytech switches...

As i understand, the patch (cacti-strip_quotes.patch (c) user) works if i use cmd.php as a poller only... what about cactid?

I could do it and then you could test? Would you like cmd.php or cactid?

Gimme a cactid fix

Thanks.

PS: while i was writing this msg, TheWitness has got an avatar =P

[TheWitness]

The current Cactid does this as well. We are planning another release soon to correct some issues though.

TheWitness

[en3my]

Mmm? ... I've upgraded to the latest cacti/cactid yesterday... :/

CACTID 0.8.6f Copyright 2002-2005 by The Cacti Group

Cacti 0.8.6g + all the patches from cacti.net

===

cacti.log(i have a "generic oid template" data source what tries to fetch the value using single oid, .1.3.6.1.2.1.2.2.1.10.26):
...
11/17/2005 01:19:10 PM - CACTID: Poller[0] Host[52] DS[830] WARNING: Result from SNMP not valid. Partial Result: Wrong Type (should b...
..

... and the output of "snmpwalk -v 2c -c xxx qqq.www.www.eee .1.3.6.1.2.1.2.2.1.10.26":
IF-MIB::ifInOctets.26 = Wrong Type (should be Counter32): Counter64: 1067884768808

PS: Thanks for superfast reply!