One of my IP aliases stops responding to pings and TCP connects

[Marco Paganini]

Hello all,

This is a weird one. I have one server configured with MC/SG. Everything runs fine except for this tiny little problem I'm having: One of the 'virtual' IPs (actually, the second one on lan2, a fddi card) stops responding to ICMP requests and TCP connections.

The weird part comes from the fact that it is completely intermittent. Sometimes, it works absolutely find. Sometimes, you can ping but not telnet. And sometimes you can't ping or telnet.

A traceroute -s using this address as the source address always fails when the problem is happening.

If I stop/start the package, or even do ifalias del IP and ifalias addmask mask ip on the interface, everything comes back to normal (and stays like that for hours).

I have a rather constant message on the syslog saying "fddi dropped interrupt" or something similar (sorry, I'll be near the server tomorrow).

This behavior began after we installed MC/SG. We have it running on other servers without any problems. Lanscan, netstat and ifconfig reveal nothing interesting about the interface at the moment of the problem. The original interface (not the alias) works without any problems.

Any ideas?

Regards,
Paga

[Steven Sim Kok Leong]

Hi,

What about a linkloop test? Can you run a linkloop test to the MAC address of an adjacent system within the same subnet?

If this works, then it is a layer-3 issue. If this doesn't work, it is a layer-2 (MAC) or perhaps even a hardware problem. If that's the case,

Try to eliminate the problem by
1) replacing the network cable.
2) use a different network card.
3) connect to a different switch port.

Hope this helps. Regards.

Steven Sim Kok Leong
Brainbench MVP for Unix Admin
http://www.brainbench.com

[John Waller · Accepted solution]

At the time when lan2:2 fails does the address allocated to lan2:1 still work. If so then this eliminates any chance of a card or cable problem. I've seen symptoms like this when their has been a duplicate IP address on my network. Modern network designs with switches don't help this as sometimes they keep directing trafic to the wrong port even after the offending machine with the duplicate address has been turned off.
I also would question using ifalias to delete and re-add the IP address. I don't know how this differs to the cmmodnet command ServiceGuard uses

[Charles Harris]

Hi,

Just a quick thought, have you ever received a source quench messages from ping when the connection goes 'funny'? It may be the box preventing fresh connections and stopping icmp pings etc because there is too much, network wise, going on. NDD can be used on HP-UX11.00 to check or NETTUNE on 10.20.....

Hope this helps....


-ChaZ-

[Marco Paganini]

Hello Folks,

I think I was too quick on the post. :) This morning I had some extra minutes and I gave some attention to this problem. I checked the obvious and it was there: A duplicate IP address (guess the OS, guess? guess???). Well, then I came to check the forums and somebody suggested just that.

Aren't the forums a cool tool? :)

Regards,
Paga