net-snmp

On Fri, 2006-02-24 at 16:53 +0100, Malcolm Townsend wrote:
> I note also there is an interface.c file in both /mibII and
> in /mibgroup/if-mib/data_access.

Forget about the mibII code - that's the original version,
inherited from the CMU package.  The 'if-mib' subtree is
the new re-implementation, which tries to separate the
architecture-specific aspects of gathering the underlying
data (data_access), from the architecture-independent aspects
of reporting this data via SNMP.

If you can put together a suitable solaris version of the
data_access files, then everything else Should Just Work.
That's the aim, certainly.

I'm sure Robert can provide more details if need be.

Dave

On Fri, 24 Feb 2006 16:53:52 +0100 Malcolm wrote:
MT> -I note that the kernel statistics are read in the file kernel_sunos5.c via
MT> kstats. ifInoctets is derived from ipackets (amultiplied by 308 to get
MT> octets presumably) . So I therefore gather for the ifHCinoctets counter
MT> ipackets64 would have to be read.

If such a structure can be read, great. But it isn't necessary. The code takes
care of using a 32 bit counter and watching for it to wrap, and updating the
hc counters.

MT> -I note also there is an interface.c file in both /mibII and in
MT> /mibgroup/if-mib/data_access. The comparable FreeBSD patches modifies
MT> interface.c, so presumably the same must be done for solaris.

as Dave said, ignore mibII, except as a reference for how the old code got the
32bit values. Really, you could just cut-n-paste that code into the right
places in the new solaris if-mib/data_access functions, and it should just
work.

MT> Could someone possibly point me in a general direction ?  I would be
MT> grateful for any tid-bits.

If you have an IRC client, come pester me in the net-snmp irc channel. You'll
get much more timelier responses there.

-- 
NOTE: messages sent directly to me, instead of the lists, will be deleted
      unless they are requests for paid consulting services.

Robert Story; NET-SNMP Junkie
Support: <http://www.net-snmp.org/> <irc://irc.freenode.net/#net-snmp>
Archive: <http://sourceforge.net/mailarchive/forum.php?forum=net-snmp-coders>

You are lost in a twisty maze of little standards, all different.

Hi,

Thank you for your response. Unfortunately I do not have access to IRC.

So what you are saying is that:

-I can just cut and paste all the code related to Solaris from
/mibII/interfaces.c and place this in a new file under /if-mib/data_access
--> interface_solaris.c.

-The pasted code recuperates statistics (by calling kernel_sunos5.c).

-Have the interface_solaris.c file use netsnmp_arch_interface_container_loa=
d
(like in solaris_linux.c).

-place the statistics recuperated by the pasted /mibII/interfaces.c code an=
d
place it into the netsnmp_interface_entry structure.

Is that the way to proceed ? Or not ?

Cheers,

Malcolm


On 2/24/06, Robert Story <rstory@...> wrote:
>
> On Fri, 24 Feb 2006 16:53:52 +0100 Malcolm wrote:
> MT> -I note that the kernel statistics are read in the file
> kernel_sunos5.c via
> MT> kstats. ifInoctets is derived from ipackets (amultiplied by 308 to ge=
t
> MT> octets presumably) . So I therefore gather for the ifHCinoctets
> counter
> MT> ipackets64 would have to be read.
>
> If such a structure can be read, great. But it isn't necessary. The code
> takes
> care of using a 32 bit counter and watching for it to wrap, and updating
> the
> hc counters.
>
> MT> -I note also there is an interface.c file in both /mibII and in
> MT> /mibgroup/if-mib/data_access. The comparable FreeBSD patches modifies
> MT> interface.c, so presumably the same must be done for solaris.
>
> as Dave said, ignore mibII, except as a reference for how the old code go=
t
> the
> 32bit values. Really, you could just cut-n-paste that code into the right
> places in the new solaris if-mib/data_access functions, and it should jus=
t
> work.
>
> MT> Could someone possibly point me in a general direction ?  I would be
> MT> grateful for any tid-bits.
>
> If you have an IRC client, come pester me in the net-snmp irc channel.
> You'll
> get much more timelier responses there.
>
> --
> NOTE: messages sent directly to me, instead of the lists, will be deleted
>      unless they are requests for paid consulting services.
>
> Robert Story; NET-SNMP Junkie
> Support: <http://www.net-snmp.org/> <irc://irc.freenode.net/#net-snmp>
> Archive: <
> http://sourceforge.net/mailarchive/forum.php?forum=3Dnet-snmp-coders>
>
> You are lost in a twisty maze of little standards, all different.
>

Hi,

I have question,

Would anyone know who calls var_ifEntry in Solaris ?  i.e, this function
makes I imagine requests to kernel_sunos5.c to find out the actual
statistics values (I guess).

I'm trying to follow my intuition, and copy the pertinent
/mibII/interfaces.c code into a new /data_access/interface_solaris.c file
(see below)

then again I could be completely on the wrong planet...


On 3/9/06, Malcolm Townsend <sootynsweep@...> wrote:
>
>  Hi,
>
> Thank you for your response. Unfortunately I do not have access to IRC.
>
> So what you are saying is that:
>
> -I can just cut and paste all the code related to Solaris from
> /mibII/interfaces.c and place this in a new file under /if-mib/data_acces=
s
> --> interface_solaris.c.
>
> -The pasted code recuperates statistics (by calling kernel_sunos5.c).
>
> -Have the interface_solaris.c file use
> netsnmp_arch_interface_container_load (like in solaris_linux.c).
>
> -place the statistics recuperated by the pasted /mibII/interfaces.c code
> and place it into the netsnmp_interface_entry structure.
>
> Is that the way to proceed ? Or not ?
>
> Cheers,
>
> Malcolm
>
>
>  On 2/24/06, Robert Story <rstory@...> wrote:
>
> > On Fri, 24 Feb 2006 16:53:52 +0100 Malcolm wrote:
> > MT> -I note that the kernel statistics are read in the file
> > kernel_sunos5.c via
> > MT> kstats. ifInoctets is derived from ipackets (amultiplied by 308 to
> > get
> > MT> octets presumably) . So I therefore gather for the ifHCinoctets
> > counter
> > MT> ipackets64 would have to be read.
> >
> > If such a structure can be read, great. But it isn't necessary. The cod=
e
> > takes
> > care of using a 32 bit counter and watching for it to wrap, and updatin=
g
> > the
> > hc counters.
> >
> > MT> -I note also there is an interface.c file in both /mibII and in
> > MT> /mibgroup/if-mib/data_access. The comparable FreeBSD patches
> > modifies
> > MT> interface.c, so presumably the same must be done for solaris.
> >
> > as Dave said, ignore mibII, except as a reference for how the old code
> > got the
> > 32bit values. Really, you could just cut-n-paste that code into the
> > right
> > places in the new solaris if-mib/data_access functions, and it should
> > just
> > work.
> >
> > MT> Could someone possibly point me in a general direction ?  I would b=
e
> > MT> grateful for any tid-bits.
> >
> > If you have an IRC client, come pester me in the net-snmp irc channel.
> > You'll
> > get much more timelier responses there.
> >
> > --
> > NOTE: messages sent directly to me, instead of the lists, will be
> > deleted
> >      unless they are requests for paid consulting services.
> >
> > Robert Story; NET-SNMP Junkie
> > Support: <http://www.net-snmp.org/> <irc://irc.freenode.net/#net-snmp>
> > Archive: <http://sourceforge.net/mailarchive/forum.php?forum=3Dnet-snmp=
-coders
> > >
> >
> > You are lost in a twisty maze of little standards, all different.
> >
>
>
>

On Thu, 2006-03-09 at 20:07 +0100, Malcolm Townsend wrote:
> Would anyone know who calls var_ifEntry in Solaris ?

It's called by the main agent driving code, when a request
is received for a MIB object within the interfaces tree.

If you want to follow the flow of control, it basically
starts with 'handle_snmp_packet' (in snmp_agent.c), then
netsnmp_handle_request, netsnmp_handle_pdu, handle_var_requests
which calls the relevant handler chain.
  In the case of the traditional interfaces.c code, this
ends up invoking the 'old_api' helper, which calls the
var_ifEntry routine.



> I'm trying to follow my intuition, and copy the
> pertinent /mibII/interfaces.c code into a
> new /data_access/interface_solaris.c file

That sounds about right, yes.

Dave

Hi,

I have several questions:

We have used the getkstat function to retrieve the ipackets and opackets
values.

1) From previous answers:

"The code takes care of using a 32 bit counter and watching for it to wrap,
and updating the
hc counters."

Q: So do we just enter the 32 bit value that is retrieved using kstat into

entry -> stats.ibytes.low

What do I have to put in entry -> stats.ibytes.high if this is the case ?

2) The values are only updated after roughly 27 seconds. You have to wait
this amount of time before a new value is posted. Sending a snmpget

Q1: Where can one modify this value modified ?

Q2: Why is this value roughly 25 seconds ?

Q3: It seems that snmp_api.c (/snmplib directory) funtion
snmp_sess_select_info deals with this. What do we need to do to make this
quicker ??
 3) Why do we have to multiply 308 by ipackets to get ifInOctets and if
InHCoctets?
Thank you for response !

Malcolm




On 3/13/06, Dave Shield <D.T.Shield@...> wrote:
>
> On Thu, 2006-03-09 at 20:07 +0100, Malcolm Townsend wrote:
> > Would anyone know who calls var_ifEntry in Solaris ?
>
> It's called by the main agent driving code, when a request
> is received for a MIB object within the interfaces tree.
>
> If you want to follow the flow of control, it basically
> starts with 'handle_snmp_packet' (in snmp_agent.c), then
> netsnmp_handle_request, netsnmp_handle_pdu, handle_var_requests
> which calls the relevant handler chain.
> In the case of the traditional interfaces.c code, this
> ends up invoking the 'old_api' helper, which calls the
> var_ifEntry routine.
>
>
>
> > I'm trying to follow my intuition, and copy the
> > pertinent /mibII/interfaces.c code into a
> > new /data_access/interface_solaris.c file
>
> That sounds about right, yes.
>
> Dave
>

On Tue, 21 Mar 2006 11:52:32 +0100 Malcolm wrote:
MT> We have used the getkstat function to retrieve the ipackets and opackets
MT> values.
MT> 
MT> 1) From previous answers:
MT> 
MT> "The code takes care of using a 32 bit counter and watching for it to wrap,
MT> and updating the
MT> hc counters."
MT> 
MT> Q: So do we just enter the 32 bit value that is retrieved using kstat into
MT> entry -> stats.ibytes.low

correct.

MT> What do I have to put in entry -> stats.ibytes.high if this is the case ?

Nothing. It should have been initialized to 0.

MT> 2) The values are only updated after roughly 27 seconds. You have to wait
MT> this amount of time before a new value is posted.

The getkstat values, or snmpd value?

MT> Q1: Where can one modify this value modified ?
MT> 
MT> Q2: Why is this value roughly 25 seconds ?

If it's the snmpd value, because that should be frequent enough that even a
gigabit interface wouldn't have time to wrap a 32bit counter.

MT> Q3: It seems that snmp_api.c (/snmplib directory) funtion
MT> snmp_sess_select_info deals with this. What do we need to do to make this
MT> quicker ??

ifTable_data_access.h:#define IFTABLE_CACHE_TIMEOUT   30

Keep in mind that you are increasing the load  on the system by more frequent
polling.

MT>  3) Why do we have to multiply 308 by ipackets to get ifInOctets and if
MT> InHCoctets?

I have no earthly idea. That's old code, and I think maybe the byte counts
weren't available, and so they are fudging it based on packets and a guess at
average packet size. If you can get real values, go for it.

-- 
NOTE: messages sent directly to me, instead of the lists, will be deleted
      unless they are requests for paid consulting services.

Robert Story; NET-SNMP Junkie
Support: <http://www.net-snmp.org/> <irc://irc.freenode.net/#net-snmp>
Archive: <http://sourceforge.net/mailarchive/forum.php?forum=net-snmp-coders>

You are lost in a twisty maze of little standards, all different.