Recent changes to feature-requests

#104 AgentX should pass through GETBULK requests

Bill Fenner — Fri, 19 Apr 2013 19:31:33 -0000

1) should be a non-issue, because you can use the SearchRange to scope the AgentX-GetBulk to exactly the registered range. (It's almost as if they thought about this problem! ;-)

2) could also be handled with SearchRange, by including the beginning of the more-specific registration as the ending OID, right?

#104 AgentX should pass through GETBULK requests

Wes Hardaker — Fri, 19 Apr 2013 14:15:47 -0000

Well, the WWW part of the comment does indicate it was me.

My recollection is that there are a few issues. It is likely safe to do some of the time, like when the response is all you need for the outgoing packet to finish.

But consider a few other cases:

1) the response runs beyond the area covered by the agentx registration boundary (eg, the subagent returns stuff beyond range). Each varbind in the response needs to be checked and possibly thrown out if it went too far.

2) the case where another subagent has registered a lower level (more specific) oid. We have code to deal with this now, but it needs to be done for each varbind as well. Although my recollection is that the registration ranges are split into two when subagents register a lower level OID. But it's been too long since I've looked at that code.

So I do think some "fixing" likely needs to be done, though not what that function is doing.

#104 AgentX should pass through GETBULK requests

Bill Fenner — Fri, 19 Apr 2013 13:26:39 -0000

The AgentX master code has been updated to be a handler, and even specifies HANDLER_CAN_GETBULK. However, it translates getbulk into getnext still:

    case MODE_GETBULK:         /* WWWXXX */
        pdu = snmp_pdu_create(AGENTX_MSG_GETNEXT);
        break;

and

        if (cache->reqinfo->mode == MODE_GETBULK)
            netsnmp_bulk_to_next_fix_requests(requests);

It's tempting to see what happens if the master agent just creates GETBULK instead of GETNEXT, and remove the netsnmp_bulk_to_next_fix_requests call.

(Does anyone know why this was done? Maybe because old subagents didn't implement GETBULK because the master agent didn't implement GETBULK?)

#175 lifecycle of snmpset ?

youssef — Mon, 25 Feb 2013 10:54:23 -0000

any help pleaaaze !!!

lifecycle of snmpset ?

youssef — Mon, 25 Feb 2013 09:59:49 -0000

Hi,
When I run snmpset commands to insert the data in the MIB,
I could clearly see the data with snmpwalk:

for example :

snmpset

snmpset -v 2c -c private 127.0.0.1 .1.3.6.1.4.1.4890.1.1.2.\"mtdcsun0\" i 89

PING-TABLE-MIB::entier."mtdcsun0" = INTEGER: 89

snmpset -v 2c -c private 127.0.0.1 .1.3.6.1.4.1.4890.1.1.4.\"nom_du_serveur\" D 0.652

PING-TABLE-MIB::reels."nom_du_serveur" = INTEGER: 89

snmpwalk -v 2c -c private 127.0.0.1 .1.3.6.1.4.1.4890.1.1

PING-TABLE-MIB::entier."mtdcsun0" = INTEGER: 89
PING-TABLE-MIB::reels."nom_du_serveur" = INTEGER: 0.652000

But the problem arises when I reboot the server I can not find any data in the MIB(by snmpwalk) "empty data base".

Is there a way to not lose data with snmpset every time I restart the server ?

any help would be greatly appreciated

#173 UDP ping support for DISMAN-PING

Bill Fenner — Fri, 08 Feb 2013 19:24:31 -0000

Do you actually have systems that implement the UDP echo service? Since it enables "ping-pong" and "Fraggle" style attacks, systems have generally shipped with it disabled by default (or not even implemented) since the late '90s.

There is no place to provide SNMP credentials with which to do a pingSnmpQuery, although I suppose you could have some defaults.

Similarly, there's no place to provide a TCP port number for a pingTcpConnectionAttempt.

I think the other values were fundamentally things that the remops working group thought might be nice but didn't really think through completely, and there hasn't been a great demand for implementing them.

#174 bulk snmptranslate usage

Bill Fenner — Wed, 06 Feb 2013 21:59:50 -0000

See https://sourceforge.net/p/net-snmp/patches/1248/ .

#154 snmtrapd and AgentX (at run-time)

Niels Baggesen — Wed, 06 Feb 2013 21:51:33 -0000

status: open --> accepted
assigned_to: Niels Baggesen
milestone: -->

#154 snmtrapd and AgentX (at run-time)

Bill Fenner — Sat, 02 Feb 2013 20:29:17 -0000

See https://sourceforge.net/p/net-snmp/patches/1247/ . I even picked the same command line option.

#167 notificationLogMib module at snmpd without a running snmptra

Niels Baggesen — Fri, 25 Jan 2013 04:56:07 -0000

status: open --> closed
milestone: -->