net-snmp-coders — Discussion of internal project coding (Please use this OR -users, NOT both!)

Re: DISKIO MIB questions, still :-)

[Stan B. <st...@aw...>]

On Tue Nov 14 12:35:57 2000 Wes Hardaker wrote...
>
>
>Stan> Here is what I get when I run snmpwalk from another machien, so
>Stan> that I can see the numeric OID's, which I need to put into the
>Stan> mrtg config file.
>
>If you run snmpwalk with "-On" on the command line, it should print
>the OIDs numerically for you.
>

	Ecelent! thanks for the tip.

-- 
Stan Brown     st...@aw...                                    843-745-3154
Charleston SC.
-- 
Windows 98: n.
	useless extension to a minor patch release for 32-bit extensions and
	a graphical shell for a 16-bit patch to an 8-bit operating system
	originally coded for a 4-bit microprocessor, written by a 2-bit 
	company that can't stand for 1 bit of competition.
-
(c) 2000 Stan Brown.  Redistribution via the Microsoft Network is prohibited.

RE: agentx initialization fails with exit(1)

[Szaboky, G. (George) <sz...@lu...>]

Yes, I have seen this behavior also. The code tries to setup the connection
with the master agent. If this fails exit() is called. For what it's worth,
I agree with Wes. The subagent should only exit optionally. In my app I
would like see the failure returned and sleep some period of time and try to
establish the connection later. 

George

-----Original Message-----
From: Wes Hardaker [mailto:wjh...@uc...]
Sent: Tuesday, November 14, 2000 11:46 AM
To: Lawrence Greenfield
Cc: net...@li...
Subject: Re: agentx initialization fails with exit(1)


>>>>> On Mon, 13 Nov 2000 15:44:14 -0500, Lawrence Greenfield
<le...@an...> said:

Lawrence> I've noticed that the AgentX API will fail with an exit(1)
Lawrence> inside of init_agent()---really subagent_pre_init().

Bad Dave, Bad Dave!

Lawrence> Since my application (the Cyrus IMAP server) will be able to
Lawrence> run just dandy without SNMP instrumentation, I'd like to log
Lawrence> an error and keep going.

It shouldn't be exiting, so yes we should fix this.  More
appropriately, it probably should only be optionally exiting at the
most.

Dave?  Your opinions?

-- 
Wes Hardaker
Please mail all replies to net...@li...
_______________________________________________
Net-snmp-coders mailing list
Net...@li...
http://lists.sourceforge.net/mailman/listinfo/net-snmp-coders

Re: agentx initialization fails with exit(1)

[Wes H. <wjh...@uc...>]

>>>>> On Mon, 13 Nov 2000 15:44:14 -0500, Lawrence Greenfield <le...@an...> said:

Lawrence> I've noticed that the AgentX API will fail with an exit(1)
Lawrence> inside of init_agent()---really subagent_pre_init().

Bad Dave, Bad Dave!

Lawrence> Since my application (the Cyrus IMAP server) will be able to
Lawrence> run just dandy without SNMP instrumentation, I'd like to log
Lawrence> an error and keep going.

It shouldn't be exiting, so yes we should fix this.  More
appropriately, it probably should only be optionally exiting at the
most.

Dave?  Your opinions?

-- 
Wes Hardaker
Please mail all replies to net...@li...

Re: Newbie needs help(URGENT) - MIBs: what for/why?

[Wes H. <wjh...@uc...>]

>>>>> On Mon, 13 Nov 2000 12:46:32 +0100, Klasema Terry <ter...@wi...> said:

Klasema> I'm totally new in using SNMP. I want to write an app that
Klasema> reads some infos from a network device. When I program an
Klasema> action (e.g. click on button should read a specific value
Klasema> from the network device) I already have to know the OID of
Klasema> that value. So I only need the MIB to get this OID once. But
Klasema> therefore a piece of paper with the OIDs and the datatypes
Klasema> would be enough.  Why do I then have to load a MIB? Well, I
Klasema> know that hardcodeing the OIDs in my program isn't a good
Klasema> way. So do I have to load a MIB and what are the advantages?
Klasema> I would be very glad if someone could help before I do a
Klasema> wrong design of my app!

1) The MIBs merely allow you to translate an OID into a textual
   representation of it (which is frequently more meaningful to human
   eyes).

2) You might start by reading the tutorials, which discuss this kind
   of information.  http://net-snmp.sourceforge.net/tutorial (see the
   snmptranslate command for example).

-- 
Wes Hardaker
Please mail all replies to net...@li...

Re: DISKIO MIB questions, still :-)

[Wes H. <wjh...@uc...>]

Stan> Here is what I get when I run snmpwalk from another machien, so
Stan> that I can see the numeric OID's, which I need to put into the
Stan> mrtg config file.

If you run snmpwalk with "-On" on the command line, it should print
the OIDs numerically for you.

-- 
Wes Hardaker
Please mail all replies to net...@li...

SMUX - snmpset

[Kapil V. <Ka...@ne...>]

Hi
    I am using zebra 0.89/ and ucd 4.1.2/ and using smux for BGP MIB
support. I have bgpd and snmpd running on a machine and bgp peer running on
another machine. when I do an snmpwalk, i get 19 variables only. But when i
try to do an snmpset on any table-entry variable in the BGP MIB(which has a
read-write permission), the snmpd is killed, smux connection is closed, core
file is generated and there is no reponse.  
  When I ran gdb on core, it gave the me following output indicating where
core is generated:

Core was generated by `./snmpd'.
Program terminated with signal 11, Segmentation fault.
Reading symbols from /lib/libdb.so.2...done.
Reading symbols from /usr/lib/libz.so.1...done.
Reading symbols from /lib/libm.so.6...done.
Reading symbols from /lib/libc.so.6...done.
Reading symbols from /lib/ld-linux.so.2...done.
Reading symbols from /lib/libnss_files.so.2...done.
Reading symbols from /usr/lib/gconv/ISO8859-1.so...done.
#0  0x0 in ?? ()
(gdb) where
#0  0x0 in ?? ()
#1  0x804b71a in handle_var_list (asp=0x80f82b0) at snmp_agent.c:576
#2  0x804b5ab in handle_next_pass (asp=0x80f82b0) at snmp_agent.c:495
#3  0x804b3fc in handle_snmp_packet (operation=1, session=0x80fdea8,
    reqid=1024378492, pdu=0x80fe760, magic=0x0) at snmp_agent.c:383
#4  0x806fff6 in _sess_read (sessp=0x80f84f0, fdset=0xbffff8a4)
    at snmp_api.c:3582
#5  0x8070050 in snmp_sess_read (sessp=0x80f84f0, fdset=0xbffff8a4)
    at snmp_api.c:3605
#6  0x806f5e9 in snmp_read (fdset=0xbffff8a4) at snmp_api.c:3272
#7  0x804ab1b in receive () at snmpd.c:668
#8  0x804a7f5 in main (argc=1, argv=0xbffffbb4) at snmpd.c:584


can somebody tell me, why the smux connection gets closed or why the snmpd
process gets killed, when i try to "set" any variable, which has read-write
permissions.  Is this related to problems in smux configuration or is it an
implementation bug?

thanx in advance
kapil

Net Brahma Technologies
Internext Networking Software
A Microland Group Venture

ka...@ne...
www.netbrahma.com

Re: Pluggable transport domains via register_readfd-like API?

[John N. <jb...@ca...>]

Hi,

  as a matter of interest, I now have a preliminary version of snmplib 
working that supports pluggable transport layers.  At the moment I have 
implemented (and partially tested) the standard UDP domain, the TCP and Unix 
domains which UCD-SNMP has traditionally supported (but which I don't think 
are official standard SNMP transport domains) and an AAL5 PVC domain (which 
is why I was doing this in the first place).  The latter is only supported 
on Linux; it would be pretty easy to extend this to Solaris as well but I 
don't have the relevant bits and pieces to hand at the moment.  It should be 
very easy to slot in other transports as long as they support a socket-like 
interface (so IPX support should be easy enough I believe).

  This hasn't been too difficult so far.  The remaining bits and pieces that 
need working on are to do with traps, address-based filtering and some 
miscellaneous other places where the library/agent/apps currently assume an 
IP transport.  Finally there is the small matter of portability (hopefully 
this is actually helped by the more structured nature of the code).  
Disappointingly, it hasn't made all that much difference to the size of 
snmp_api.c -- it's about 100 lines shorter, that's all.

  If anyone is interested, I can email patches, but for now I won't post 
anything to the sourceforge page. I expect to clean up the remaining stuff 
over the next few days.

  What's the current ETA of 4.2?

John
-- 
// Dr. John Naylon // Senior Software Engineer // Cambridge Broadband Ltd.
// The future of broadband wireless access  // www.cambridgebroadband.com

test... ignore this.

[Wes H. <wjh...@uc...>]

-- 
"Ninjas aren't dangerous.  They're more afraid of you than you are of them."

agentx initialization fails with exit(1)

[Lawrence G. <le...@an...>]

I've noticed that the AgentX API will fail with an exit(1) inside of
init_agent()---really subagent_pre_init().

Since my application (the Cyrus IMAP server) will be able to run just
dandy without SNMP instrumentation, I'd like to log an error and keep
going.  Would a patch to change

void init_agent(const char *app)

to 

/* returns non-zero on error; no further snmp calls should be made
   unless this returns success */
int init_agent(const char *app)

be accepted?

Thanks,
Larry

Re: SNMP.pm

[<gm...@no...>]

On what do you base the conclusion that "it will only work on an NT"?
did you get a binary distribution for Win32?

more details please.

-GSM

%% Regarding SNMP.pm; you wrote:

  f> Hi
  f> It appears that the UCD SNMP agent compiled and did a make install fine on Solaris
  f> 7.  Unfortunately 
  f> the version of SNMP.pm that came bundled with it will only work on an NT platform.
  f>  Why they 
  f> bundled this with a Unix SNMP agent is beyond me.  I can't seem to find a non
  f> RPM version of your 
  f> SNMP.pm on the web.  Could you send me a zip of it withe the make install stuff.
  f>  I would greatly 
  f> appreciate it.

  f> Thanks

  f> Fred Johnson



-- 
G.S. Marzot                        email: gm...@no...
Nortel Networks                    voice: (978)288-3990
600 Tech Park  M/S E65-60-405      pager: (800)409-6080 (40...@sk...)
Billerica, MA  01821                 fax: (978)670-8145

Re: Master agent core dump

[<ni...@ba...>]

On Mon, 13 Nov 2000 12:19:32 +0100 Didier Chabanol <did...@fe...>
wrote:
>Due to the memory leak with sub agent coabitation, we are using a
>single master agent for our test.
>
>The problem seems to be in this line :
>The memory adress pointed by (void *) ((char *) cachep->cache_addr +
>(cachep->cache_last_found * entrysize)) points out of table limit.

In what file? There are very many files in this project, and several thousand
lines of code, so could you point a little more precisely? Btw, in what version
of ucd-snmp?

/Niels

-- 
Get your firstname@lastname email for FREE at http://Nameplanet.com/?su

Re: Master agent core dump

[Didier C. <did...@fe...>]

Oupssss... Sorry

file is : kernel_sunos5.c
OS : Sun 5.7
version : 4.1.1
masteragent supporting MIB II and host Mib.
No sub agent


ni...@ba... a écrit :

> On Mon, 13 Nov 2000 12:19:32 +0100 Didier Chabanol <did...@fe...>
>
> wrote:
>
> >Due to the memory leak with sub agent coabitation, we are using a
>
> >single master agent for our test.
>
> >
>
> >The problem seems to be in this line :
>
> >The memory adress pointed by (void *) ((char *) cachep->cache_addr +
>
> >(cachep->cache_last_found * entrysize)) points out of table limit.
>
> In what file? There are very many files in this project, and several thousand
>
> lines of code, so could you point a little more precisely? Btw, in what version
>
> of ucd-snmp?
>
> /Niels
>
> --
> Get your firstname@lastname email for FREE at http://Nameplanet.com/?su

Newbie needs help(URGENT) - MIBs: what for/why?

[Klasema T. <ter...@wi...>]

Hi all!
I'm totally new in using SNMP. I want to write an app that reads some infos
from a network device. When I program an action (e.g. click on button should
read a specific value from the network device) I already have to know the
OID of that value. So I only need the MIB to get this OID once. But
therefore a piece of paper with the OIDs and the datatypes would be enough.
Why do I then have to load a MIB? Well, I know that hardcodeing the OIDs in
my program isn't a good way. So do I have to load a MIB and what are the
advantages?
I would be very glad if someone could help before I do a wrong design of my
app!
 
Thank you very much!
Regards
 
              Terry

Master agent core dump

[Didier C. <did...@fe...>]

Due to the memory leak with sub agent coabitation, we are using a
single master agent for our test.

The problem seems to be in this line :
The memory adress pointed by (void *) ((char *) cachep->cache_addr +
(cachep->cache_last_found * entrysize)) points out of table limit.


    if (cache_valid) {
      /* Entry is valid, let's try to find a match */
      if (req_type == GET_NEXT) {
        result = getentry(req_type,
                          (void *) ((char *) cachep->cache_addr +
(cachep->cache_last_found * entrysize)),
                          cachep->cache_length - (cachep->cache_last_found *
entrysize),
                          entrysize, &ep, comp, arg);
      } else {
        result = getentry(req_type, cachep->cache_addr, cachep->cache_length,
     entrysize, &ep, comp, arg);
      }
    }

Dave Shield a écrit :

> Didier> Our validation department is making some load test on this particular
> Didier> master agent.
> Didier> After 20 hours the master agent crash.
>
> Mike> Try again, this time please use the current version at
> Mike> ftp://ucd-snmp.ucdavis.edu/beta/ucd-snmp-4.2.pre1.tar.gz
>
> The other thought that springs to mind, is to ask whether
> you are testing a single-agent configuration, or whether this
> is using a master/slave AgentX setup.
>
>   If the latter, then it is quite possible that the (known)
> memory leaks in the AgentX implementation may result in problems
> when the system is hammered in this way.
>   I hope to be able to look at this again before the 4.2 release,
> but as always, this is subject to pressures of (paid) work.
>
> Dave

Re: Agentx question on how to register for a new index

[Dave S. <D.T...@cs...>]

> How do I request an index

There's a series of routines

	register_{string,int,oid}_index

that can be used for this.
They each take the OID to register under (complete with length),
plus the "obvious" value parameter.

There are special values ANY_{STRING,INTEGER,OID}_INDEX to
request an arbitrary value.   Not that this high-level API
does not support requesting a "new" value - you'd need to
tweak the code, or use the lower level 'register_index' routine

Note that this particular facility has not been widely used
so far, so there may well be bugs lurking there!


> and once I have an index, register the row?

That's done much as usual.  Simply use the index you've
been given to form the OID(s) to register.

  If you want to register a row slice, you'll probably
need to use the routine "register_mib_range".  See the file
'agent_registry.c' for the (long!) list of default parameters.
We probably need a cleaner interface for this.

  Again, this probably hasn't been tested as thoroughly as
the rest of the agent, so be prepared for problems.  :-)


Dave

Re: A first look at diskio for HP-UX

[Dave S. <D.T...@cs...>]

Didier> Our validation department is making some load test on this particular
Didier> master agent.
Didier> After 20 hours the master agent crash.

Mike> Try again, this time please use the current version at
Mike> ftp://ucd-snmp.ucdavis.edu/beta/ucd-snmp-4.2.pre1.tar.gz

The other thought that springs to mind, is to ask whether
you are testing a single-agent configuration, or whether this
is using a master/slave AgentX setup.

  If the latter, then it is quite possible that the (known)
memory leaks in the AgentX implementation may result in problems
when the system is hammered in this way.
  I hope to be able to look at this again before the 4.2 release,
but as always, this is subject to pressures of (paid) work.

Dave

Re: snmp

[Bert D. <dri...@pl...>]

On Sun, 12 Nov 2000, lucy chapa wrote:

> We need some help with the UCD SNMP module.  The module was installed and 
> configured on a unix box running solaris and we were able to do some command 
> line gets and was able to walk through the mib tree however when the program 
> was run the errors received were "Cannot  find module (IP-MIB): At line 0 in 
> (none), Cannot find module (TCP-MIB): At line 0 in (none), etc, etc"

As you surmised, this is due to the libraries inability to find the MIB
files.

> It looked like the module was unable to locate the mibs we changed the 
> .profile file to go to a specific directory and located the rfc1850, 
> rfc1406, rfc1407 mibs and put them there.

This is almost certainly the wrong approach. If you built and installed
the entire project, it should have installed correct MIBs into
/usr/local/share/snmp/mibs (or another place if you specified a
different prefix).

>  Now I am getting the following 
> errors (just a sampling):
> Cannot find module (IP-MIB): At line 12 in (none)

If you sucessfully built and installed the project, then continued with
the PERL bits, the only thing I can think of is that something is wrong
in your environment. I'd suggest checking that MIBS and MIBDIRS are
unset.

You should not have to replace any of the MIBs that come with NET-SNMP.

You'll have to do a "make;make install" in the top net-snmp build
directory before trying to test the Perl bits, IIRC.

Cheers,

				-- Bert

Bert Driehuis -- dri...@pl... -- +31-20-3116119
If the only tool you've got is an axe, every problem looks like fun!

snmp

[lucy c. <luc...@ho...>]

We need some help with the UCD SNMP module.  The module was installed and 
configured on a unix box running solaris and we were able to do some command 
line gets and was able to walk through the mib tree however when the program 
was run the errors received were "Cannot  find module (IP-MIB): At line 0 in 
(none), Cannot find module (TCP-MIB): At line 0 in (none), etc, etc"

It looked like the module was unable to locate the mibs we changed the 
.profile file to go to a specific directory and located the rfc1850, 
rfc1406, rfc1407 mibs and put them there.  Now I am getting the following 
errors (just a sampling):
-------------------------------------------
PERL_DL_NONLAZY=1 /usr/local/bin/perl -Iblib/arch -Iblib/lib 
-I/usr/local/lib/pt
t/libtest..........."my" variable $value masks earlier declaration in same 
stat.
"my" variable @oid_names masks earlier declaration in same statement at 
blib/li.
Cannot find module (IP-MIB): At line 12 in (none)
Cannot find module (IF-MIB): At line 12 in (none)
Cannot find module (TCP-MIB): At line 12 in (none)
Cannot find module (UDP-MIB): At line 12 in (none)
Cannot find module (SNMPv2-MIB): At line 12 in (none)
Cannot find module (SNMPv2-SMI): At line 12 in (none)
Cannot find module (UCD-SNMP-MIB): At line 12 in (none)
Cannot find module (UCD-DEMO-MIB): At line 12 in (none)
Cannot find module (SNMP-VIEW-BASED-ACM-MIB): At line 12 in (none)
Cannot find module (SNMP-COMMUNITY-MIB): At line 12 in (none)
Cannot find module (SNMP-FRAMEWORK-MIB): At line 12 in (none)
Cannot find module (SNMP-MPD-MIB): At line 12 in (none)
Cannot find module (SNMP-USER-BASED-SM-MIB): At line 12 in (none)
Cannot find module (SNMPv2-SMI): At line 18 in 
/usr/local/lib/MIBS/rfc1850.mib
Cannot find module (SNMPv2-SMI-v1): At line 20 in 
/usr/local/lib/MIBS/rfc1850.mb
Bad operator (Jan): At line 39 in /usr/local/lib/MIBS/rfc1850.mib
----------------------------------------------------

We thought that the mibs might be the wrong version.  Any help would be 
appreciated.

Lucy Chapa
luc...@ho...
_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.

Share information about yourself, create your own public profile at 
http://profiles.msn.com.

RE: 4.2.pre1 dumps core on Solaris 2.6

[<ni...@ba...>]

On Fri, 10 Nov 2000 16:36:57 +0100 RADMAN Stefan <Ste...@CT...>
wrote:
>
>Niels> It actually failed when it tried to generate the error message! APply
>the
>Niels> following patch
>
>Does that mean you were not able to reproduce the problem ?

It means that I did reproduce it, but your analysis was wrong.

>Thanks for the patch - basically it worked but I had 3 problems:
>(1) third hunk failed - guess that's why I dont get the "(Node: (top) -> "
>output but just the oid I requested

Oops, sorry about that, I did the patch against something more recent than
4.2.pre1 which may have affected it.

/Niels

-- 
Get your firstname@lastname email for FREE at http://Nameplanet.com/?su

Agentx question on how to register for a new index

[Szaboky, G. (George) <sz...@lu...>]

Hi, 

I am developing an agentx subagent that uses a private mib. I want to
register for a row in a table. How do I request an index and once I have an
index, register the row?

George

Re: Problems building latest Perl/SNMP

[Terminator r. <kar...@el...>]

That would be very good.  I actually had done a 'make install' of the
NetSNMP code, but of course it was into '/tmp/net-snmp-test' so perl
didn't find it.

Sorry for the confusion,

	-r

On 10 Nov 2000, G. S. Marzot wrote:
 > perl Makefile.PL NET-SNMP-PATH=/usr/local

-- 
O----O     Karl "rat" Schilke -- kar...@el... -- (360)816-3837
 \Oo/   
==\/==             Caution: Fruit Filling May Be Extremely Hot.

Re: Problems building latest Perl/SNMP

[<gm...@no...>]

I am not sure it would make sense to build and install Perl/SNMP without
having net-snmp installed already. I do think that Makefile.PL needs a
way to specify explicitly where the install path is for net-snmp because
people often have a stable (old) release and a new one that they are
trying out.

something like

perl Makefile.PL NET-SNMP-PATH=/usr/local

-GSM

-- 
G.S. Marzot                        email: gm...@no...
Nortel Networks                    voice: (978)288-3990
600 Tech Park  M/S E65-60-405      pager: (800)409-6080 (40...@sk...)
Billerica, MA  01821                 fax: (978)670-8145

Re: help required!

[Wes H. <wjh...@uc...>]

>>>>> On Thu, 9 Nov 2000 20:52:19 -0800 (PST), jaya Bharathi <mbh...@ya...> said:

jaya> I have newly moved to snmp programming side.  I have written a
jaya> MIB file.  I need to compile and use that mib file in my
jaya> project. I need to perform all the operations on the mib
jaya> like(get,set etc). How do I compile the MIB file using ucd snmp
jaya> tool kit.  Please advise me on this.

There is a tutorial on this subject at
http://net-snmp.sourceforge.net/tutorial/toolkit.  Additionally, you
should read the AGENT.txt file found in the ucd-snmp source-code's tar
ball.

-- 
Wes Hardaker
Please mail all replies to net...@li...

RE: 4.2.pre1 dumps core on Solaris 2.6

[RADMAN S. <Ste...@ct...>]

Niels> It actually failed when it tried to generate the error message! APply
the
Niels> following patch

Does that mean you were not able to reproduce the problem ?

Thanks for the patch - basically it worked but I had 3 problems:
(1) third hunk failed - guess that's why I dont get the "(Node: (top) -> "
output but just the oid I requested
(2) had to add '#include "tools.h"' to mib.c to define SNMP_FREE

Thanks 
Stefan

Here is the (new) patch that worked for me (pure 4.2.pre1):
*** snmplib/mib.c.orig  Fri Nov 10 14:03:57 2000
--- snmplib/mib.c       Fri Nov 10 15:00:21 2000
***************
*** 79,84 ****
--- 79,85 ----
  #include "read_config.h"
  #include "snmp_debug.h"
  #include "default_store.h"
+ #include "tools.h"

  static struct tree * _sprint_objid(char *buf, oid *objid, size_t
objidlen);
  static char *uptimeString (u_long, char *);
***************
*** 1252,1257 ****
--- 1253,1259 ----
      tree_top = (struct tree *)calloc(1,sizeof(struct tree));
      /* XX error check ? */
      if (tree_top) {
+        tree_top->label = strdup("(top)");
          tree_top->child_list = tree_head;
      }
  }
***************
*** 1260,1266 ****
  shutdown_mib (void)
  {
      unload_all_mibs();
!     free(tree_top); tree_top = NULL;
      tree_head = NULL;
      Mib = NULL;
      free(Prefix); Prefix = NULL;
--- 1262,1271 ----
  shutdown_mib (void)
  {
      unload_all_mibs();
!     if (tree_top) {
!        SNMP_FREE(tree_top->label);
!        free(tree_top); tree_top = NULL;
!     }
      tree_head = NULL;
      Mib = NULL;
      free(Prefix); Prefix = NULL;

snmptrapd fork and closing files

[Wes H. <wjh...@uc...>]

Is there any reason why anyone can think that the following is
necessary in the snmptrapd code:

                /* Close all unnecessary file descriptors */
		fdnum = getdtablesize();
		for ( fd = (STDERR_FILENO + 1); fd < fdnum; fd++ )
			close(fd);

There shouldn't be any other file descriptors, since snmptrapd isn't
itself forked.

(I'm trying to deal with a snmp_logging problem that stems from the
logging file descriptor getting closed)

-- 
"Ninjas aren't dangerous.  They're more afraid of you than you are of them."