ICMP ping problem

[mmcgurty]

I have an interesting problem using Linux. From testing we've conducted, it appears that Red Hat 8.0, Red Hat 9.0, and SuSe 7.3 all have the same issue. We use Linux to monitor critical servers, routers, and switches using Nagios. We've come a long way with this, to the point that we've got voice notifications on the phone after hours of critical devices that effect our customer. In any case, it seems that in all these version of Linux whenever the polled network is lost it can never again reach the network after coming back up using ICMP. For example:

Nagios on Linux is on 192.168.40.0/21, it's checking 192.168.138.0/24 over frame-relay. The frame-relay circuit drops, 192.168.138.0/24 is now unreachable. Services are unreachable, Nagios on Linux generates a message to our SMS phones after so many missed polls. Now, when the frame-relay circuit comes back up you can ICMP ping 192.168.138.1/21 (default gateway, in our case a router) but anything beyond that is unreachable with ICMP ping. However, telnet or anything other than ICMP is reachable.

The only way to get it working again is to ping with certain options enabled (-i I think) or ping from the specific interface e0 on the Nagios Linux server, or reboot.

The default routes, gateways, etc match all of our Windows machines that work just fine. We are thinking this is a problem that just exists with Linux in general.

Any ideas would be appreciated. Thank you.

 

[edemiere]

When this happens do an ifconfig -a - is the interface marked "up" or "down"?

Let me know what it says...

Cheers!

 

[msteder]

You might want to talk to the Nagios tech support people. Do they use the linux ping? They probably have their own ping utility, like a multiping or something proprietary.