NetView documentation gives the advice to
put a caching name server on your Netview Server. I would take that
a step further and 'fix' troublesome nameserver configurations.
1. Run named on your NetView server
2. Using an external process, discover your network and bring in the
names
3. Using a script, process the NetView topology database and generate your
own DNS zone files. Put in the SNMP address of the device into the name
lookup table. Put ALL the interfaces into the reverse lookup files so that
they map to the device name NetView is using.
4. Update your resolv.conf to use your local DNS first, then network
DNS second. This will give you good solid NetView DNS support with
very little CPU costs. New devices coming in will get their name from the
network DNS. Their name will be configured into the local DNS the
next time you run your script to process the NetView database.
This drastically reduces CPU loads from large host files. The
nice thing is that once you have your script written, it is very close to
painless for maintaining the names.
Monday, March 08, 2004 3:08
PM To: nv-l AT lists.us.ibm DOT com cc: nv-l AT lists.us.ibm DOT com,
owner-nv-l AT lists.us.ibm DOT com From: "Christopher J Petrina"
<cjp8 AT meadwestvaco DOT com> Subject: RE: [nv-l] seed
file
OUr issue with names is
that we have a central database that has all the delvices within the network in
it (Lotus Notes DB). Not all these devices are in DNS. They have
added names for every interface on every device into this Lotus Notes DB in no
particular fashion for denoting which interfaces belong to which device.
The naming scheme is not exactly consistent. Because this is a large
database of all the devices it is what I haev to work with. and we also have the
same issues as others about HSRP ISDN backups and all that. My main
question was in how netview deals with the multiple entries of names for the
same device. Thank you for all the input.
|
| Francois Le Hir
<flehir AT ca.ibm DOT com> Sent
by: owner-nv-l AT lists.us.ibm DOT com
03/08/2004 03:51 PM Please respond to nv-l
|
To:
nv-l AT lists.us.ibm DOT com cc:
Subject: RE: [nv-l] seed
file |
I tried the host file but it gave me more trouble than
anything else as the name resolution is a lot heavier on the CPU if the host
file is big. The main reason is that the Netview server was also a syslog
server and syslogd not being multithreaded, it started eating 100% of one cpu
and dropping traffic. Maybe if you don't have other name resolution intensive
process it would work but it all depends on the size of the host
file.
Also I wonder what others do for the name resolution of HSRP
interface ? The way I do is to resolve the ip to the name of the primary
router but it lead to frequent problems when somehow Netview elect an HSRP
interface to be the "SNMP address" for a device.
Salutations, /
Regards,
Francois Le Hir Network Projects & Consulting
Services IBM Global Services Phone: (514) 205 6695
"Evans, Bill"
<Bill.Evans AT hq DOT do
e.gov>
To Sent by:
"'nv-l AT lists.us.ibm DOT com'"
owner-nv-l@lists. <nv-l AT lists.us.ibm DOT com>
us.ibm.com
cc
Subject
03/08/2004 03:21
RE: [nv-l] seed file
PM
Please
respond to
nv-l
Another important item is to use only one name in DNS
for each router. All the interfaces on that device should resolve to
the same name. The name should resolve to the loopback address for the
router.
This gives you the best functioning of NetView although there
is a continuing argument from many who want a unique DNS name for each
interface for other reasons. I find the device name qualified by the
interface name (e.g. Router.Serial1/0) gives the uniqueness I need for
interfaces. When necessary I override the DNS with a local hosts file to
achieve this. In my present situation the DNS has only the loopback
address and I use hosts to resolve the other names.
Bill
Evans
-----Original Message----- From: Paul
[mailto:pstroud AT bellsouth DOT net] Sent: Monday, March 08, 2004 2:05 PM To:
nv-l AT lists.us.ibm DOT com Subject: Re: [nv-l] seed file
Chris, Only
use one IP for each device, you are causing netmon to do a lot of additional
and unecessary cycles. Let me explain:
netmon reads the entries from
the seedfile, pings the node, if it responds, it is added to the database and
an SNMP poll in scheduled. When the device is polled via SNMP all the
interfaces are discovered and finally the full device is added to the
map.
If you put multiple IPs in the seedfile for the same device it
will be added and polled mutliple times, one for each ip in the seedfile.
This is uneccesary as netmon will already discover it via SNMP. That
and you might confuse things. Ie. netmon discovers on interface and adds
it and finds another interface and
adds it before the first SNMP poll is completed. Now it must delete one of
the devices(as we know they are on the same device) and fixup the correct
device. This is all uneccesary processing.
The best practice is to
add a single interface from each
device.
Paul
Christopher J Petrina
wrote:
> > Greetings all, > > UNIX netview.
Using a seefile forl imited discovery. In the seedfile > I
have multiple entries (multiple IP's) for the same device, ie(every > ip
interface of a single router) is in the seedfile. When netview >
runs through the seedfile it finds the first IP of a device when it >
comes to the second IP of that device what does netview do. Each
IP > interface also has an entryo in DNS as well. Does netview
change the > name of the device if it finds another name. Also once
it has > initially found it, and then polls the device (SNMP) what name
will it > chose for the device in netview, and why does it sometimes
change the > name of the device. > > > Chris
Petrina
________________________________________________________________________ This
email has been scanned for all viruses by the MessageLabs
SkyScan service._______________________________________________________________
|