Re: Troubleshooting connection loss (continued)

On Thu, 08 Nov 2007 23:03:02 GMT, Allen Weiner wrote:
Bit Twister wrote:

I believe ntp runs only occasionally.

Very true if clock is close to time server.

I don't have automatic email checks.

I had no idea if you had someting like thunderbird up. Default is
check every 10 minutes.


While my connection is up, the Gkrellm eth0 monitor never goes
blank, even if I'm just browsing local files on my HDD. A partial
explanation, from the other thread, is that every 15 seconds, Verizon is
probing to see if I'm running a server.

Yea, saw that. but I never see those in my Verizon router log. You would think
they would check for servers on us FiOs users.


What you posted above, did not show us that fact.

"service network restart" clears the routing table and then hangs.

Yes, but, while down, the proof would be
route -n > down.txt
ifconfig >> down.txt
and include down.txt in your reply

1. Connection loss.
2. I issue route -n. Result is same as before connection loss.

I'll take your word for it. :-)


avahi-daemon 0:off 1:off 2:off 3:on 4:on 5:off 6:off

I boot into runlevel 5. The above result shows that avahi-daemon is not
activated for runlevel 5.

Ok, I guess I'll need to add code to get runlevel. :-D


========== head -15 /etc/hosts ===========
192.168.1.1 gateway


That bites, you should have a local host entry.
I suggest one for the node, if you would give it a node name.

Novice question: What is the rationale for having a local host entry?

There are apps which want to commicate and they just want to do it on
the local host (127.0.0.1)

When I originally configured for static IP, I got the DNS server
addresses from the router. Primary: 68.237.161.12. Secondary:
71.250.0.12. Isn't everything in "12" Verizon?

Make sure, host ip_here_2_check

Besides, as long as I can
browse the web, DNS must be working even if the servers aren't from Verizon.

Very true, but what if some cracker manages to put their DNS servers
in there. :-(

I was worried, because of the myhome.westell.com hostname being issued to your
node via dhcp.


Now your /etc/hosts should have something like:
127.0.0.1 localhost
192.168.1.1 gateway
192.168.1.47 darkstar.home.invalid darkstar

Novice comment: I don't understand the rationale for making this change.

Need the 127. entry for local communications.
I left the gateway line.
The darkstar.home.invalid darkstar was so there was an entry for your
static ip which matched your node name.
That ip is normally used by your desktop manager for when it
uses your node name for ip resolution.


In the other thread you suggested 192.168.1.147,

Yep, that was to tell me if dhcp gave you an address or you were realy
were using a static value. When I indicated 140 in this thread, it
was to help keep anyone from confusing .147 with .47. Next time I'll
choose .150 :)

and that's what I used at first. But I had a setback.

My PC is dual-boot: Fedora 7 and Windows/ME. Nowadays I mostly run
Fedora, but at least once a week I run Windows/ME.

I would think one OS set dhcp and other OS set static should not
matter as long as ip address is different.
I do it all the time. First install uses dhcp and some time later I
pick a static ip. My XP Home ran dhcp until the next Second Tuesday
when I booted for updates.

I ran Fedora with static IP for several days with no problem.

Past the normal failure date???

During
that time I did not run Windows/ME. When I eventually ran Windows/ME I
also configured that for static IP. It ran OK, but the domain I selected
on the DNS configuration screen is not what is recommended.

If static, I would have set actual verizon dns values found in router.
But, if you wanted you can use the router's dns vaule (192.168.1.1)

A few days later, Fedora would not boot.
It hung in boot on "starting sendmail".

Yep, it may have not been able to lookup node name in /etc/hosts.
It may have hung because it wanted to look up the upline
relay/smart host to send email still in the mail queue and network was
not up.

I booted back into Windows/ME. It could not locate Google.

If WinME could not ping 72.14.207.99 (google.com) or 208.101.56.232
(www.google.com) then connection was still down.

After switching Fedora and Windows back to dynamic IP, everything was OK.

I would set both static with verizon dns values, power off everything,
wait 1 minute by the wall clock, powerup modem, let leds settle, power
up pc.

If you switched them to the original .47 then, the router still thinks
it handed out the lease and it is being used.

As an FYI. I answered yes to a XP Home Windows Update nic driver and I could
no longer use the nic under linux. :-(

So, I've made a second attempt with static IP. I thought I'd play it
safe and use the same IP address that DHCP was giving me.

And that is why I asked for 147. We have to break the dhcp server/client
out of the loop to see which end of the connection is causing your
drop out problem.

I used the GUI.

Ok, good, I had booted my fedora 7 install and found there is another
eth0 config file in a default/ directory awhile ago. No idea which one
is being used to control the nic.

Freaking mlocate cron script does not seem to update the locate database. I
installed updatates and need to go back and boot fc7, update the
locate database and see if locate eth0 shows all the config files when logged
in as root.

I did check, and FC7 GUI does not have a selection to allow setting
PEERDNS=no in nic config file.

If you have about 10 gig of free space, I can get you setup on
Mandriva Linux and we can rule out Fedora as your problem. :-P


Something else to play with the next time your connection drops.
Before you do the service network restart, I want you to unplug the
ethernet to pc cable from the modem or nic.
Wait at least 30 seconds by the wall clock, plug it in and do the
serice network restart and see if network comes up.

Once in a great while when playing with my network, I have had to
restart my firewall after restarting the network.

.

Relevant Pages

  • Re: Help with initial small org AD setup convention when using DMZ network
    ... Consider using Dynamic DNS internally (aka Active Directory Integrated ... > firewall which then connects the public IP dmz network to a private IP ... > domain name for such subnets based on the nearest airport code, ... > servers to serve acme.com names for external users. ...
    (microsoft.public.win2000.active_directory)
  • Re: Dynamic DNS, DNS Records & Scavenging
    ... There are two DHCP ... Both the servers are set to update A and PTR records for clients. ... DNS however this doesn't seem to be the case. ... DNSUpdateProxy and when i did this i saw my test laptop register its ...
    (microsoft.public.windows.server.dns)
  • Help with initial small org AD setup convention when using DMZ network
    ... firewall which then connects the public IP dmz network to a private IP ... domain name for such subnets based on the nearest airport code, ... Yahoo to manage my externally-visible DNS names on the acme.com domain. ... and servers that use this domain, ...
    (microsoft.public.win2000.active_directory)
  • Re: network has gone down again, and I cannot figure out why
    ... search dns: dns.asm.bellsouth.net ... try against the secondary DNS server address. ... up DHCP on this end and now I can email. ... DHCP just configures your side of the network with the data that the ...
    (Fedora)
  • Re: Domain Logon Speed
    ... Once in the new offices I certainly will be using a "formal" network ... NICs, DHCP and all the other things suggested in this ... you must specify that all servers and workstations specify ...
    (microsoft.public.windows.server.sbs)