Discovery Guided by a seedfile

After performing a few manual discoveries, you will develop the necessary confidence and data to automate the process. Armed with a list of routers in the management domain, you can create the seedfile, delete the object, topology, and map database, and restart the NNM daemons. Since this process is intended to be repeatable, it’s common to delete the database multiple times until all the configuration files, including the seedfile, are just right and discovery proceeds the way you intend.

Remember to place the seedfile in a safe and stable location independent of the NNM installation tree, such as /opt/config/seedfile. Assign ownership and permissions to the seedfile as appropriate to your local management processes.

To delete the NNM object, topology, and map database (this assumes the flat file database of NNM 6.x and later), there are two options. To keep the old database around, perform these steps:

• stop the daemons with $OV_BIN/ovstop
  
  
• rename the openview directory with mv openview openview.old. Pre-NNM 6.1. delete old event log files with rm $OV_LOG/xnmevents.*. for NNM 6.1 cd $OV_DB/eventdb ; rm -rf $OV_DB/eventdb/*/*
  
  
• clear the SNMP cache with $OV_BIN/xnmsnmpconf -clearCache
  
  
• start ovwdb with $OV_BIN/ovstart ovwdb
  
  
• execute the command $OV_BIN/ovw -fields
  
  
• start the daemons with $OV_BIN/ovstart
  
  

To remove the database instead of saving it, just delete the entire subdirectory (rm openview/*/*) instead of renaming it in the above steps. Note that alarms may be deleted.

The process of deleting the database and starting over again may be necessary because the seedfile is usually incomplete at first. It may contain routers outside the intended management domain or be missing routers in the intended management domain. As you test the new seedfile, new problems will emerge, necessitating another round of discovery. A final database scrub and rediscovery is needed to validate the final seedfile.

Note that if you merely want to add a router to the seedfile, there’s no need to scratch the database. Just stop and start netmon to force it to read the seedfile again. If you want to delete a router from the seedfile, and if it’s easy to delete the unintentionally discovered topology by hand, then do so, and stop and start netmon again.

What do you do if the seedfile is accidentally erased and there is no backup? You may go back to your notes and retype the file. After all, the file is probably only 50-100 lines long. Is there an automatic way, however, to regenerate the list from the NNM database? Why not use ovtopodump -f filtername to print out all the routers, where filtername is a filter defined in the $OV_CONF/C/filters file that correctly matches routers? This list will include routers that have unmanaged subnets attached. Using such routers in the seedfile will expand the management domain one layer of subnets and routers outside the current one. Use the Internet submap for guidance and trim the extraneous routers from the seedfile.

Note that subnets that are initially discovered from seedfile routers are managed by default. Subnets later added to these routers will be unmanaged by default. For a non-seedfile router that is autodiscovered afterwards, NNM leaves all subnets initially unmanaged except for the already-managed subnet(s) on which the router has interfaces.

Discovery driven by a seedfile is very fast most of the time. An exception is during quiet periods on the network when ARP caches have drained and many workstations are turned off. Under such conditions, autodiscovery can be painfully slow. Sometimes it’s helpful to include the jump start parameter “-J” in netmon.lrf. This allows netmon to instruct a remote system with an HP SNMP agent to transmit an ICMP echo request to its IP broadcast address, forcing all active machines on the subnet to answer with an ICMP echo reply. This fills the ARP cache of the remote system, which netmon then harvests.

Note that if the NNM system acts as a management station which is not intended to do autodiscovery on its own, the seedfile should contain only the names of the associated NNM collection stations.