Topic: Dropped Internet connection 2...

[steve lewis]

This is my situation, very perplexing.
------------------

Problem: e-smith box loses Internet connectivity at what appears to be regular intervals. Internal LAN is not affected.

Hardware: PC server, AMD K6-2 500Mhz, 256 M ram, 20G IDE drive, 2- AOpen Real-Tek 8139 NICs.

e-smith versions: this has occurred with fresh installs of  versions; 5.1.2, 5.5, 5.6b7 ,patched and unpatched.

ISP: Cox Communications, residential DHCP based, cable connectivity. The use of username /password is not in use. They do not use BOOTP style MAC address binding. I called tech support to confirm that their  system was not experiencing any problems.

Log Files: Within the “messages” log from the Server Manager, my server appears to request a new DHCP address from the DHCP server from Cox Cable Internet. However the address is not given out, a release does not take place therefore complete negotiation process fails.

Workaround: A Bug report for this issue, ISP DHCP response corrupts resolver configuration,  or one very similar has been recognized  and is at http://e-smith.org/bugs/index.php3?op=showBug&bugID=67

I use the  workaround below from the bugs page:

# drop the external link - note therefore that you can't perform the
# fix remotely using this procedure

  /sbin/ifdown eth1

# Replace /etc/resolv.conf with the correct contents

  /sbin/e-smith/expand-template /etc/resolv.conf


# Now bring the external interface back up
  /sbin/ifup eth1

Troubleshooting: When this began I was using a fresh install of 5.6b7. After the first occurance I rebooted and didn’t think much of it. The second time, rebooted and checke the bugs page where I found a similar DHCP on the external interface issue. I applied the workaround with success.

Then occurred a third time, so this time I replaced the NICs. Approximately 30 minutes later the issue happened again.

I reinstalled with 5.12 on the same hardware. Issue was not resolved, it happened again. The workaround resolved a new DHCP address and allowed Internet connectivity without a reboot.

I applied the latest patches for 5.1.2, yet the occurrences continued.

So, I reinstalled with 5.5, which was the current permanent fix for the, ISP DHCP response corrupts resolver configuration , bug.

Did not work. Then I applied all of the 5.5 patches including the non-Blades RPMs.

Did not work. All of the above and the issue was not resolved.

So now I am using a Floppy Disc based Linux router/firewall. This is the link, http://bbiagent.net/en/index.html .

Ok, I am using all the same hardware except the OS has changed, BBIagent floppy router/firewall. I have had NO problems using BBIagent router on the same hardware. Been up more than 18 hours with no interruptions.

[Bill Talcott]

steve lewis wrote:
>
> Log Files: Within the “messages” log from the Server Manager,
> my server appears to request a new DHCP address from the DHCP
> server from Cox Cable Internet. However the address is not
> given out, a release does not take place therefore complete
> negotiation process fails.

It sounds like you've discovered your own problem right there. It sounds like the SME is doing a DHCP request for an IP, and isn't getting it. I don't know exactly what the problem is, but it sounds like it's right there.

> Workaround: A Bug report for this issue, ISP DHCP response
> corrupts resolver configuration,  or one very similar has
> been recognized  and is at
> http://e-smith.org/bugs/index.php3?op=showBug&bugID=67
>
> I use the  workaround below from the bugs page:
>
> # drop the external link - note therefore that you can't
> perform the
> # fix remotely using this procedure
>
>   /sbin/ifdown eth1
>
> # Replace /etc/resolv.conf with the correct contents
>
>   /sbin/e-smith/expand-template /etc/resolv.conf
>
>
> # Now bring the external interface back up
>   /sbin/ifup eth1

This bug involves having the ISP's DHCP response corrupt the resolv.conf file. To fix it, you rebuild the file from the templates. I'm guessing it worked in your case because it involves taking down and restarting the interface, not actually anything to do with the resolv.conf/DHCP bug.

I'd say you should start with the bad DHCP renewal requests. Is the server making renewal requests when it shouldn't? Why isn't the ISP responding properly to the requests? I don't have any specific information, but this appears to be where you should start looking.

[steve lewis]

Hi,

Thanks for the response. I think it definately is trying to renew when it shouldn't. Cox Comm. has their lease set for 3hours, and in the past have I kept the same IP for several months. From the logs, which I will post later this eve, my server requests around every 30-45 minutes. If the DHCP server is initiating the negotiation I think it should show in the logs.

The floppy based router/firewall that I am using in the interim allows me to release and renew very easily with a button click, and I can watch the negotiation.

At least the floppy router/firewall has no problems releasing, renewing and keeping the IP it gets using the same hardware, it is the same machine.

What are the Linux commands to release and renew an IP address?

Just frustrating cause Cox Comm. says nothing wrong with their DHCP servers and not that knowledgable to sleuth this one out quickly, but will get there. And, never had any problems, now 3 versions do the same thing.

Thanks,

Steve Lewis

[Arkman]

To release/'kill' the lease:
/sbin/dhcpcd eth1 -k

To renew:
/sbin/dhcpcd eth1 -n

Substitute the correct interface if you have them swapped.