LMS 2.2 - Bakup archive for 3000 nodes

[gbusson]

Hi,

using LMS 2.2 RME 3.5. We have 1500 routers and 1500 switchs define.

Every day we plan a backup rchive configuration at 9 pm but we can see that backup is never finished at 6 am.

At this time, archive backup is based on the Config Retrieval Schedule.

Is using SNMP Poller schedule a best method ? What are the difference ?

It seems that LMS make a backup archive for all the nodes defined. Is it possible to sp?cify which nodes to backup (using LMS or external scripts)

Thanks

Gilles

[akemp]

The SNMP poller option is the best one to pursue IMHO just due to the number of devices. Normally you would only want to collect the delta (change) in the configurations instead of just collecting everything each day. This is more true when the device complexity increases and the configuration gets longer. Its one thing to archive 1,500 1725 configurations verses a mixture of GSR's, 7600's, 6513's etc.

[gbusson]

what is IHMO ?

I configured RME to use SNMP Poller. But it doesn't seem to work fine. Not all my devices are checked. I have modifed a configuration last day and this night this device whas not checked.

And in archive status, i can see that all my device are not checked.

[akemp]

Sorry for our non-English conversant members. IMHO "in my honest opinion" . Just click on the failed "No. of devices" link under archive management and it will give you a failure code as to why each failed under the "description" field . Most are pretty easy to figure out with the description. Missing credentials or unreachable devices are to most common culprits.

My earlier post also alluded to the difficulties of actually collecting the configurations but did not enumerate all the possibilities for problems: Not enough bandwidth to the device, device is busy, too many devices for a single RME server to poll in an hour, RME server capibilities, etc. The list is legion.

I manage some 3,000 active devices across the US with 2 Ciscoworks LMS 2.5.1 clusters... not single a single processor windows server, but 4-4 processor Sun boxes with 8GB of memory which helps a bunch. It still takes me a good 6 hours for a complete inventory collection job ...and thats not with a 100% success rate either.

[gbusson]

If i click on failed, i've no device. I set a network trace. And in the trace i can see that CW2K doesn't make request for all devices. In your message you write 'too many devices for a single RME ....in a hour' Is there a restriction ?

In the network trace, i can see that the archive process run for only 45 minutes.

[akemp]

What sort of credentials and protocols do you have configured to use for configuation retrieval ?