nv-l
[Top] [All Lists]

Re: [nv-l] Polling devices that are loadhosts'd into object database

To: nv-l@lists.us.ibm.com
Subject: Re: [nv-l] Polling devices that are loadhosts'd into object database
From: "Christopher J Petrina" <cjp8@meadwestvaco.com>
Date: Wed, 7 Jan 2004 17:39:22 -0500
Delivery-date: Wed, 07 Jan 2004 22:46:32 +0000
Envelope-to: nv-l-archive@lists.skills-1st.co.uk
Reply-to: nv-l@lists.us.ibm.com
Sender: owner-nv-l@lists.us.ibm.com

I guess an explanation of how we do things is in order.

We have a database of all devices on the network.  CLients, Servers, Routers, Switches, and many others.   We dump that file into a comma delimited file and it is FTP'd over to the Netview box.  a Third-party software grabs it, and creates a seedfile from that data.   at the same time it executes a loadhosts command on each line of the file to load it into the objects database.  This is done nightly.   The next time it receives a file it checks the file against the previous one if there are new entries in the new one it strips only those out and then adds them via loadhosts to the objects database and appends to the seedfile.   I understand what you mean by the redundancy of a seedfile and using loadhosts.  Our devices do support SNMP, however, we are in the process of making sure they are all configured correctly.  

I could be insane or not but I have a memory in my head that someone told me support level or I read somewhere that anything discovered via a SEEDFILE will NEVER use ICMP for polling.  Am I crazy?




Stephen Hochstetler <shochste@us.ibm.com>
Sent by: owner-nv-l@lists.us.ibm.com

01/07/2004 05:15 PM
Please respond to nv-l

       
        To:        nv-l@lists.us.ibm.com
        cc:        
        Subject:        Re: [nv-l] Polling devices that are loadhosts'd into object database



This is all a bit interesting.

1. Do your devices support SNMP? Have you updated NetView with the right community strings? If so, then only one IP address PER device would need to be in the seed file. NetView will walk the MIB of that device and discover the other IP addresses from the interface table. You are doing overkill by including all IP addresses...just one per device is needed with discovery turned off.

2. If you are using loadhosts, then you don't need a seedfile for discovery. It is redundant. Again, if the devices support SNMP, the loadhost will load one interface into NetView, NetView will find other interfaces from the MIB interface table. With discovery off, NetView will not read the ARP cache and find other nodes from this device.

3. Using a basic seedfile has NOTHING to do with ICMP or SNMP status polling. However, you can build a seedfile with specific syntax to tell NetView to only use SNMP polling for specific devices. It needs additional characters to force NetView to do that.

Summary:
-- using loadhosts is fine.....but it does hide some benefits of using NetView for you. When you run loadhosts, you give the subnet of the interface. With discovery from NetView, he reads the subnet mask as it is configured in the devices. This 'visual' view of your network will help you find subnet masks that are mis-configured. If you only use loadhosts, you hide that fact. I assume you are using loadhosts so you don't get a bunch of the desktops and other items you don't want to discover. There is another approach that would satisfy this with potentially better results.

Many people do not want NetView to discover their whole network at once due to the load on your network. But you really have 3 choices: 1) discovery off 2) discovery on 3) discovery of backbone.

I would recommend that you start NetView (blank map) with discovery of backbone. Then I would use your application to build a seed file of just your servers you want to manage. With the seedfile, no need to also do loadhosts. Save your loadhosts for devices that don't want to come into NetView. You can try loadhosts then walk the device's MIB to determine why he has problems. Many times it will be due to a bad subnet mask. Then I would do an ovtopodump and compare that list of devices against the whole list you want to discover. What is missing would need to be investigated and could be a result of mistakes in their network configurations. Every customer I have helped has found configuration mistakes in their network due to NetView's discovery. Some minor, some not so minor.



Stephen Hochstetler shochste@us.ibm.com
International Technical Support Organization at IBM
Office - 512-838-6198 (t/l 678) FAX - 512-838-6931

http://www.redbooks.ibm.com
Inactive hide details for "Christopher J Petrina" <cjp8@meadwestvaco.com>"Christopher J Petrina" <cjp8@meadwestvaco.com>


"Christopher J Petrina" <cjp8@meadwestvaco.com>
Sent by: owner-nv-l@lists.us.ibm.com

01/07/2004 03:12 PM
Please respond to nv-l



To: nv-l@lists.us.ibm.com
cc:
Subject: [nv-l] Polling devices that are loadhosts'd into object database




Question for the masses.


ENV: Netview 7.1.4 AIX 5.1


We have a product that creates a seedfile using IP addresses. In this seedfile are router, switches, servers, etc. TO INCLUDE all the interface IPs of said routers, switches, servers, ETC.

We have Discover New nodes disabled in the setup so only what is in the seedfile will be discovered. However, the Objects database is loaded via creating a loadhosts line for each IP in the seedfile.

the loadhosts line includes a -p -P -M (with proper subnet) and then the IP address found in each line fo the seedfile.

OK heres the question. According to documentation on the AIX version of Netview a device loaded via seedfile will not be polled using ICMP only SNMPGET. Does the fact that I am doing a loadhost into the objects database negate that fact? Or am I completely insane?


Thanks

Chris Petrina
________________________________________________________________________
This email has been scanned for all viruses by the MessageLabs SkyScan
service._______________________________________________________________


<Prev in Thread] Current Thread [Next in Thread>

Archive operated by Skills 1st Ltd

See also: The NetView Web