Ok, is this possible gurus? Thanks in advance!

Ok, is this possible gurus? Thanks in advance!

Post by EL C » Wed, 17 Oct 2001 15:18:04



Ok heres a good one for all the gurus out there. At work , we have
several park servers sitting on our t1 connections.  My boss asked me
if it would be possible to find out how much bandwith some servers are
comsuming.  The servers are running linux.  No, the servers belongs to
clients, so we dont want to compromise their security or anything, but
simply wishes to find out how much bandwith they are currently using.
I know that certain manageable switches can tell you the amount of
data that is passing through a port.  I would like to know what I can
do in order to find out the amount of data going to those servers.
thanks guys!!
 
 
 

Ok, is this possible gurus? Thanks in advance!

Post by Eric P. McC » Wed, 17 Oct 2001 15:31:27



> Ok heres a good one for all the gurus out there. At work , we have
> several park servers sitting on our t1 connections.  My boss asked me
> if it would be possible to find out how much bandwith some servers are
> comsuming.

A Perl script called mrtg will do this, but you need to install an
SNMP daemon.  It will even chart the output in a reasonably pretty
graph, complete with weekly, monthly, and yearly (I think) summaries.
It works on 5-minute averages, though I believe that's configurable.

A better solution would be to watch /proc/net/dev, but then,
/proc/net/dev seems to wrap after 2^32 bytes (which I hit pretty much
every four hours now).

--

"I woke up this morning and realized what the game needed: pirates,
pimps, and gay furries."  - Rich "Lowtax" Kyanka

 
 
 

Ok, is this possible gurus? Thanks in advance!

Post by EL C » Wed, 17 Oct 2001 15:40:26


thanks!!!



>> Ok heres a good one for all the gurus out there. At work , we have
>> several park servers sitting on our t1 connections.  My boss asked me
>> if it would be possible to find out how much bandwith some servers are
>> comsuming.

>A Perl script called mrtg will do this, but you need to install an
>SNMP daemon.  It will even chart the output in a reasonably pretty
>graph, complete with weekly, monthly, and yearly (I think) summaries.
>It works on 5-minute averages, though I believe that's configurable.

>A better solution would be to watch /proc/net/dev, but then,
>/proc/net/dev seems to wrap after 2^32 bytes (which I hit pretty much
>every four hours now).

>--

>"I woke up this morning and realized what the game needed: pirates,
>pimps, and gay furries."  - Rich "Lowtax" Kyanka

 
 
 

Ok, is this possible gurus? Thanks in advance!

Post by Joe Blogg » Wed, 17 Oct 2001 17:08:06



> Ok heres a good one for all the gurus out there. At work , we have
> several park servers sitting on our t1 connections.  My boss asked me
> if it would be possible to find out how much bandwith some servers are
> comsuming.  The servers are running linux.  No, the servers belongs to
> clients, so we dont want to compromise their security or anything, but
> simply wishes to find out how much bandwith they are currently using.
> I know that certain manageable switches can tell you the amount of
> data that is passing through a port.  I would like to know what I can
> do in order to find out the amount of data going to those servers.
> thanks guys!!

If you are using Cisco switches and your servers are centrally located you
can simply run show top from the switch - which will give you your biggest
users. There are also several "nice" tools within Cisco works or NMCView
which will also give you this.. Cisco also gives you the possibility to
'span' relevant ports or geographically local vlans across to sniffers.
And failing Cisco, you should resort to using sniffers anyway. Investing
in a good sniffer package is well worth the money in future time saved
scratching your heads.
 
 
 

Ok, is this possible gurus? Thanks in advance!

Post by Dean Thompso » Wed, 17 Oct 2001 23:07:21


Hi!,


> > Ok heres a good one for all the gurus out there. At work , we have
> > several park servers sitting on our t1 connections.  My boss asked me
> > if it would be possible to find out how much bandwith some servers are
> > comsuming.

> A Perl script called mrtg will do this, but you need to install an
> SNMP daemon.  It will even chart the output in a reasonably pretty
> graph, complete with weekly, monthly, and yearly (I think) summaries.
> It works on 5-minute averages, though I believe that's configurable.

You will actually need to install the entire MRTG package to get all of the
scripts and executables that come with MRTG.  MRTG can be found at
http://www.mrtg.org and you may need to install the GD library so that it can
create the graphics for you as well.

See ya

Dean Thompson

--
+____________________________+____________________________________________+

| Bach. Computing (Hons)     | ICQ     - 45191180                         |
| PhD Student                | Office  - <Off-Campus>                     |
| School Comp.Sci & Soft.Eng | Phone   - +61 3 9903 2787 (Gen. Office)    |
| MONASH (Caulfield Campus)  | Fax     - +61 3 9903 1077                  |
| Melbourne, Australia       |                                            |
+----------------------------+--------------------------------------------+

 
 
 

Ok, is this possible gurus? Thanks in advance!

Post by EL C » Thu, 18 Oct 2001 02:07:12


THANKS GUYS!

On Tue, 16 Oct 2001 09:08:06 +0100, Joe Bloggs



>> Ok heres a good one for all the gurus out there. At work , we have
>> several park servers sitting on our t1 connections.  My boss asked me
>> if it would be possible to find out how much bandwith some servers are
>> comsuming.  The servers are running linux.  No, the servers belongs to
>> clients, so we dont want to compromise their security or anything, but
>> simply wishes to find out how much bandwith they are currently using.
>> I know that certain manageable switches can tell you the amount of
>> data that is passing through a port.  I would like to know what I can
>> do in order to find out the amount of data going to those servers.
>> thanks guys!!

>If you are using Cisco switches and your servers are centrally located you
>can simply run show top from the switch - which will give you your biggest
>users. There are also several "nice" tools within Cisco works or NMCView
>which will also give you this.. Cisco also gives you the possibility to
>'span' relevant ports or geographically local vlans across to sniffers.
>And failing Cisco, you should resort to using sniffers anyway. Investing
>in a good sniffer package is well worth the money in future time saved
>scratching your heads.

 
 
 

1. An advanced SVR4 package question (for guru's).

Hi:

Im pretty advanced when it comes to making SVR4 packages.
But here is a problem I don't know how to get around.

Im creating a package who's objects (files) belong to the user
"devel" and to the group "develgrp". When I run pkgadd on
a live O/S, the package installs fine...

The preinstall script first creates a "devel" UNIX account and
"develgrp" UNIX group; then the files are installed, the ownership
of which are specified to be devel:develgrp in the prototype file.
pkgchk gives the null response upon execution (success).

Now, I want to install this very same package during Jumpstart.
Now the scripts were coded to inherit the PKG_INSTALL_ROOT
variable (i.e. pkgadd -R ...), and know to work relative to that
base directory during a translated root.

The problem is this... the devel:develgrp account is created
in /a/etc/passwd and /a/etc/shadow, but pkgadd doesn't
refer to /a/etc/passwd and /a/etc/shadow when trying to resolve
ownership of files when it is installing them (it always refers to
/etc/passwd & /etc/shadow). So the files come up as owned
by random numbers.

Now I know that I can play around with postinstall scripts
to adjust the ownership of the files once they are installed,...
but doing so immediately corrupts the installation because
the file's ownership attributes will never match that which
is specified in the /var/sadm/install/contents file.

I also dont' want to do a chroot /a pkgadd... because that
will potentially affect package scripts (things like df -k,
begin to fail).

Finally... specifying ownerships by number in the prototype
file (as opposed to by name) causes other problems.

There must be well known way to get around this problem.

Anyone know?

Regards,
Milton

2. nscd question

3. Need a slightly advanced guru class redirection/process substitution recipe.

4. Problems with multiport board

5. Xconfig for ALG230X card NEEDED. thanks in advance.

6. Source Control Software Query

7. ATI Mach 64 - Please help me....Thank you in advance !!!

8. NAT / routed combination question

9. Can somebody describe me how to build Linux kernel, Thanks in advance.

10. Can somebody suggest some Networking Project, Thanks in advance.

11. ask for help about chroot, Thanks in advance.

12. ADVANCED PROGRAMMING QUESTION (IN LINUX BUT ANY UNIX ANSWER OK)

13. ask for help about chroot, Thanks in advance.