1720's losing connection

[jwcurtis]

I have implemented about 4 different 1720 routers. Problem is, they each need re-booted about every other day. This obviously shouldn't be happening, but it is. Is there a limit on the amout of NAT translations? The reason I ask is, when the LAN users aren't able to browse out, I am still able to telnet in from a remote location, and reload the router. Once I telnet in, I can ping LAN side address', and WAN side address', but the users aren't able to browse until I re-boot the router. I also was able to get browsing working again on one occaision, by using the clear ip nat trans * command. This hasn't fixed it except in one case, but it just seemed odd. My config is a basic one, with just a few Static NAT's for SMTP, etc, and dynamic nat with overload on one of the public IP's.

[svermill]

If memory is the culprit, you will probably be able to figure it out looking at 'sh proc mem.' I am aware of a case where nat was causing a problem in a smaller router, but that was for a large apartment building with lots of users. Adjusting the translation timeouts did the trick. But that doesn't sound like your issue. Check that memory usage out and then go from there. Could just be a leak in faulty IOS?

[brad.caldwell]

I'm having the exact same problem with one of my clients using a 1720 with WIC-1ADSL. I was told by TAC to use dsl "operating-mode ansi-dmt" to allow the WIC card to retrain in the event of line noise. This hasn't solved the problem. I don't think it's a NAT problem.

[mberrocal]

what IOS version are you using?

you could search for a NAT related bug at: http://www.cisco.com/cgi-bin/Support/Bugtool/home.pl

also when problems starts and you are still able to telnet the router check how busy is the CPU with 'show proc cpu' and memory with 'show proc mem' this could lead to a memory leak or other kind of bug. Also you could try reducing the translations timeout: 'ip nat translation timeout 5400'