See my comments after each question.
>> /etc/services contains an entry like "snmp-trap 162/udp" and
>> "nvtrapd-trap 162/udp". This specifies how the trap is sent and says
>> to use port 162 with protocol UDP. Is port 162 the UNIX domain socket?
/etc/services defines the ports on where daemons will listem for incoming
traps, not on the ones they will use to go out. 162/udp is where the trap
recipient will listen. A UNIX domain socket will not be defined in
/etc/services as it is a temporary in-box connection and uses tcp. It is a
standard UNIX method of inter-process communication.
>> When a Netview internal trap is generated how is the destination
determined?
Take netmon for example. He knows what box he is running on -- that's in
your NetView database. So he knows to open a
domain socket to trapd, who must be running on the same box. trapd can
forward traps somewhere else, but you have to tell him where in his startup
parameters (see the man page on trapd).
>> Does the trap community specification in /etc/snmpd.conf determine this
or does something else do it?
There is an entry in snmpd.conf which specifies where traps should go.
Unless you mess with it, there is only one, 127.0.0.1, which is the
loopback address of your own box. If you change it, then snmpd will send
his outgoing traps to other sources. But NetView internal traps are
incoming, so it has no effect on them.
>> Does it matter if /etc/snmpd.conf does not contain a trap community
specification? Yes, (see above) but what you put for the community name
doesn't matter much. That is what snmpd will use as the community name in
the traps he sends out. But most trap recipients don't care what that is.
NetView for example will process and display the trap regardless of the
community name. You can use the $C variable in trapd.conf to have it
displayed if you like.
>>We have written a process which also generates traps. The process uses
OVsnmpSend to send the trap PDU. Is the trap >> destination determined in
the same way as a Netview internal trap?
I would urge whoever wrote this to spend some time with the NetView
Programmer's Guide on this subject. OVsnmpSend requires you to do an
OVsnmpOpen first. This opens a session with snmpd. He sends the trap you
generate, like any other he gets, to the destination(s) specified in the
trap entry of snmpd.conf. If that is your own loopback address, 127.0.0.1
then the trap does not go out over the net (where would it go?) and I
believe that a UNIX domain socket connection is used to send to that
address, but I am not certain, as that is snmpd and not NetView.
>> I don't think we are losing any internal traps but it is an awkward
question someone has asked me which I can't answer.
Why not turn on snmpd logging? If you edit etc/snmpd.conf and change the
logging level from its default of "0" to "3" he will write a detailed log
to the file given in the other logging entry. You can then see what traps
are being sent and where they do. See the man page on snmpd for more
details, and read the comments in /etc/snmpd.conf. This can be used as the
log of your own traps. If you are worried about losing NetView traps,
don't be, unless you have daemons dying unexpectedly and that sort of
thing.
Good luck.
James Shanks
Level 3 Support
Tivoli NetView for UNIX and NT
Please note that my new id is jshanks@us.ibm.com
Robin James <robin.james@thalesatm.com>@tkg.com on 09/18/2001 09:04:38 AM
Please respond to IBM NetView Discussion <nv-l@tkg.com>
Sent by: owner-nv-l@tkg.com
To: IBM NetView Discussion <nv-l@tkg.com>
cc:
Subject: Re: [NV-L] Netview internal traps
Excuse my ignorance...
We have written a process which also generates traps. The process uses
OVsnmpSend to send the trap PDU. Is the trap destination determined in
the same way as a Netview internal trap?
Ultimately, I am trying to determine whether the traps we generate are
going to use the same mechanism as the Netview internal traps. We are
using the Netview SNMP API but is there anything we must do to ensure
our traps do not go over the network when we are using them for our own
internal purpose?
--
Robin
email: robin.james@thalesatm.com
tel: +44 (0) 1633-862020
fax: +44 (0) 1633-868313
James Shanks wrote:
>
> NetView internal traps are sent by UNIX domain socket from the sender
> (usually netmon) to trapd. They do not go out over the network. They
just
> look like they do. In fact, they never leave the box unless you forward
> them somewhere.
>
> James Shanks
> Level 3 Support
> Tivoli NetView for UNIX and NT
> Please note that my new id is jshanks@us.ibm.com
>
> Robin James <robin.james@thalesatm.com>@tkg.com on 09/18/2001 05:13:26 AM
>
> Please respond to IBM NetView Discussion <nv-l@tkg.com>
>
> Sent by: owner-nv-l@tkg.com
>
> To: NetView Discussion <nv-l@tkg.com>
> cc:
> Subject: [NV-L] Netview internal traps
>
> Netview uses internal traps. When a trap is sent over a LAN using UDP it
> can get lost. Is it possible for Netview internal traps to get lost when
> you're using a single processor computer?
>
> I don't think we are losing any internal traps but it is an awkward
> question someone has asked me which I can't answer.
>
> --
> Robin
> email: robin.james@thalesatm.com
> tel: +44 (0) 1633-862020
> fax: +44 (0) 1633-868313
> _________________________________________________________________________
> NV-L List information and Archives: http://www.tkg.com/nv-l
>
> _________________________________________________________________________
> NV-L List information and Archives: http://www.tkg.com/nv-l
_________________________________________________________________________
NV-L List information and Archives: http://www.tkg.com/nv-l
|