From OSPF to RIP

[Beebe]

I'm temporarily taking a step backwards from OSPF to RIP, because apparently as many people may have discovered, OSPF does not work properly on wireless networks.

So last night I moved from OSPF to RIP. Very smooth transition for the most part. Just a case of disable OSPF and enable RIP on each router one by one back to the NOC. I just did it step by step a chunk at a time. I discovered I could, on any router on the network, in OSPF enable redistribute RIP, and in RIP, redistribute OSPF and run both protocols at once and it would act like a gateway between OSPF and RIP and I could run rip on the further part, and OSPF on the closer part (to the NOC), and slowly transition the whole network over, chunk by chunk.

So, I'm 100% RIP now. But one problem. I got up this morning and one of the routers was not reachable. I could ping it from a war board on the same subnet/switch, but could not log in via SSH. I took my laptop to town, plugged it into the same switch and was able to log in. That was the first strange thing. Why was I able to log in from my laptop, but not from a neighboring StarOS machine?

Now I have a second board acting up. Seems like the same problem. I can ping it from a neighboring board, but can't log into it. I get

SSH Exited: Error connecting: Connection refused

It is some level of version 3 on the board I'm trying to log into, not sure exactly which. The board I'm able to ping it from is 1.3.4

Thanks,
Roger

[ripv]

FWIW the very odd time I've seen these symptops - ie being able to ping a star os box but get that connection refused error attempting to putty - has been on wrap boards when they seem to be partly crashed - a power off and back on has always rectified it for me.

[DrLove73]

I had same problem but on v2. I was told it was MAC poisoning. We went to 1.2.xb, then to 1.3.x, and problem is gone. We had static routing.

Maybe CPU load maxed out.

[Beebe]

In the past with partly crashed boards, I have been able to log in as far as putting in my username and password and then the gui partly loads, and then freezes. In which case I can use starutil -r to reboot the thing and it is back to normal.

In this case, I can't ping the board from my local machine. I can ping it from a neighboring staros board, but I can't log into it from that board, even partially.

Ahh, the penny drops...

The board was partly crashed. If this were the case when I was using OSPF, the routes would still exist and I would be able to do starutil -r to reboot the thing. As it were, I called the guy who's house it was on, and had him unplug it and plug it back in. That started it going again, and I can now log in.

OOooooh.. this is a 1.1.9 board and SNMP was running. oh, goody goody gum drops... there was a memory leak in that version if I recall correctly. So I can just turn SNMP off and I should be golden.

Awesome! Didn't even have to leave the house.

Thanks,
Roger

[tog]

When you find a loose thread on your network like this, you should take it upon yourself to sit down and login to every single one of your infrastructure units and check them. And maybe your client units, too.

You should also update any 1.1.7 - 1.1.9 units immediately. I believe 1.1.7 - 1.1.9 had a "save" bug where, whenever you saved your config, there was a small chance to corrupt your configuration and cause your board to be unbootable and require a JTAG.

If I were you, even if you aren't ready to jump to 1.3.x, I would sit down and login to every unit and check that SNMP is turned off and confirm that they are using fairly safe firmware versions.