Re: To understand the increase in size of buffer in agentx_parse() in net-snmp-5.8

2019-05-13 Thread Bart Van Assche

On 5/13/19 12:10 PM, Vijay, Anjali wrote:
Thanks for the help. Meanwhile, I had changed the buffer size back to 
1472 and it seems to be working fine like the previous version, net-snmp 
5.7.3. Do you think this can cause any serious impact?


Hi Anjali,

I think that approach is risky. It's easy for an SNMP manager to send an 
snmp get, getnext or getbulk request with multiple OIDs. Even if the 
request fits in a 1472 byte network packet, the response may exceed that 
packet size. Unless if you know from beforehand what kind of requests 
the SNMP manager(s) will send and if you know from beforehand what the 
maximum string length will be, I think it's safe to use a larger buffer 
size.


Bart.


___
Net-snmp-coders mailing list
Net-snmp-coders@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/net-snmp-coders


RE: To understand the increase in size of buffer in agentx_parse() in net-snmp-5.8

2019-05-13 Thread Vijay, Anjali
Hi Bart,

Thanks for the help. Meanwhile, I had changed the buffer size back to 1472 and 
it seems to be working fine like the previous version, net-snmp 5.7.3. Do you 
think this can cause any serious impact?

Regards,
Anjali Vijay

From: Bart Van Assche 
Sent: Sunday, April 28, 2019 3:39 AM
To: Vijay, Anjali ; net-snmp-coders@lists.sourceforge.net
Subject: Re: To understand the increase in size of buffer in agentx_parse() in 
net-snmp-5.8

On 4/25/19 5:56 PM, Bart Van Assche wrote:
On 4/25/19 8:00 AM, Vijay, Anjali wrote:
I recently started working with net-snmp-5.8. I had a query over a change that 
was made to agentx_parse() in agent/mibgroup/agentx/protocol.c. I noticed that 
the buffer of  SNMP_MAX_MSG_SIZE (1472) in net-snmp-5.7.3 has been replaced 
with the size 65536 in net-snmp-5.8.

u_char buffer[65536];

I didn't find any clarification in the ChangeLogs for this large increase of 
size. Can this size be reduced? It would be of great help if anyone can clear 
me out on this since my application has memory constraints.

Hi Anjali,

I will have a look and see whether it is possible to modify the AgentX code 
without reducing performance and without dropping support for large AgentX 
packets. BTW, the following commit increased the buffer size:

c09140c934eb ("CHANGES: snmpd: Increase maximum AgentX packet size to 64kB.")

Hi Anjali,

A candidate fix has been checked in on the v5.8 and master branches. Please 
verify.

Bart.
___
Net-snmp-coders mailing list
Net-snmp-coders@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/net-snmp-coders


Re: Update UCD-SNMP-MIB variable type

2019-05-13 Thread Niels Baggesen via Net-snmp-coders
On Mon, May 13, 2019 at 10:08:58AM +0200, Ilya Etingof wrote:
> 
> > The currelt 5.8 release (and unreleased 5.7 update) should already
> > contain objects like memTotalRealX which are defined as
> > CounterBasedGauge64
> > 
> > Unfortunately I don't have hardware to test it :-)
> 
> Just in case on-line agent emulation [1] could be helpful:
> 
> $ snmpget -v2c -c mib2dev/ucd-snmp-mib demo.snmplabs.com 
> UCD-SNMP-MIB::memTotalRealX.0
> UCD-SNMP-MIB::memTotalRealX.0 = Counter64: 18345168853242336474 kB
> 
> You can have it off-line as well [2].
> 
> 1. http://demo.snmplabs.com
> 2. http://snmplabs.com/snmpsim

Well, what is interesting is if the real agent get the real value. It
is about debugging the agent running on the physical hardware.

/Niels

-- 
Niels Baggesen - @home - Århus - Denmark - n...@users.sourceforge.net
The purpose of computing is insight, not numbers   ---   R W Hamming


___
Net-snmp-coders mailing list
Net-snmp-coders@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/net-snmp-coders


Re: Update UCD-SNMP-MIB variable type

2019-05-13 Thread Ilya Etingof


> The currelt 5.8 release (and unreleased 5.7 update) should already
> contain objects like memTotalRealX which are defined as
> CounterBasedGauge64
> 
> Unfortunately I don't have hardware to test it :-)

Just in case on-line agent emulation [1] could be helpful:

$ snmpget -v2c -c mib2dev/ucd-snmp-mib demo.snmplabs.com 
UCD-SNMP-MIB::memTotalRealX.0
UCD-SNMP-MIB::memTotalRealX.0 = Counter64: 18345168853242336474 kB

You can have it off-line as well [2].

1. http://demo.snmplabs.com
2. http://snmplabs.com/snmpsim



___
Net-snmp-coders mailing list
Net-snmp-coders@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/net-snmp-coders


Re: Update UCD-SNMP-MIB variable type

2019-05-13 Thread Niels Baggesen via Net-snmp-coders
On Mon, May 13, 2019 at 02:56:46AM -0400, Josef Ridky wrote:
> Hi folks,
> 
> I would like to know your opinion on following issue.
> 
> UCD-SNMP-MIB [1] uses INTEGER32 instead of COUNTER64 for memory OIDs which 
> limits the reporting to 2 TiB (using KiB as the base).
> Large modern servers often contain more than 2 TiB of memory leading to the 
> OIDs rolling over and reporting negative values.
> 
> How can it be solved? 
> Can we just change the variable type? (I don't think so) 
> Or can we add new variables, that will be COUNTER64 type? 
> Or can the base be set to e.g. MiB instead of KiB in configuration?

The currelt 5.8 release (and unreleased 5.7 update) should already
contain objects like memTotalRealX which are defined as
CounterBasedGauge64

Unfortunately I don't have hardware to test it :-)

/Niels

-- 
Niels Baggesen - @home - Århus - Denmark - n...@users.sourceforge.net
The purpose of computing is insight, not numbers   ---   R W Hamming


___
Net-snmp-coders mailing list
Net-snmp-coders@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/net-snmp-coders


Update UCD-SNMP-MIB variable type

2019-05-13 Thread Josef Ridky
Hi folks,

I would like to know your opinion on following issue.

UCD-SNMP-MIB [1] uses INTEGER32 instead of COUNTER64 for memory OIDs which 
limits the reporting to 2 TiB (using KiB as the base).
Large modern servers often contain more than 2 TiB of memory leading to the 
OIDs rolling over and reporting negative values.

How can it be solved? 
Can we just change the variable type? (I don't think so) 
Or can we add new variables, that will be COUNTER64 type? 
Or can the base be set to e.g. MiB instead of KiB in configuration?

[1] http://www.net-snmp.org/docs/mibs/ucdavis.html

Regards

Josef Ridky
Software Engineer
Core Services Team
Red Hat Czech, s.r.o.



___
Net-snmp-coders mailing list
Net-snmp-coders@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/net-snmp-coders