Ping etc. don't always work for me

bitpicker · 01-02-2006, 06:22 PM

I have just switched from SuSE 9.1 to Gentoo. I am encountering some strange problems with my network connection. I'm on a DSL connection through a router, and I can surf the internet without any problems whatsoever. But there are a number of strange problems which I suspect may have the same underlying reason:

When trying to emerge software from Gentoo mirrors, I frequently get no connection. Sometimes I have to try three or four different servers until I find one which replies. The servers can't be pinged at that time either; if I can ping it, it can be reached by emerge, too. So far the only strange thing is that so many servers tend to fail me, even when they worked for me hours ago and will again a couple of hours later. I put that down to holiday traffic at first, but now I'm not so sure anymore.

I emerged ntp following an article from the Gentoo wiki, and I can't get any connection to ntp servers, I always get 'address family not supported' after a while (timeout). When I use the -4 option the same thing happens, if I use -6 it happens straight away. Even the debug option -d doesn't yield any other information.

I set up a compuserve.de pop3 account with Thunderbird, and it worked well for a while. But now, during about the last four or five days, the account could not be accessed via pop3 anymore, while it can still be accessed from the web interface. I've tried different programs (mail checker, e-mail client) and they have the same problem. Surely Compuserve wouldn't drop a service for days on end?

So I went on a pinging spree, and find that I lose all packets when I ping compuserve.de or aol.com (which I tried on a whim) while I can ping google.de, google.com, linuxquestions.org and lots of other sites perfectly. Pinging about.com gives me a "Packet filtered" result, while pinging forums.about.com returns two additional duplicates of every packet...

DNS resolution works in all cases. Pinging the IP directly makes no difference.

I'm completely at a loss about this. I have to suppose that something in my new setup must be wrong, as I have never had such problems with my old system, but I have no idea what it could be. I remember similar problems from Windows times when some MRU (?) settings or whatever had to be adjusted to reach certain sites, while others worked with no problems.

Can anyone give me a hint?

Robin

bulliver · 01-03-2006, 12:54 PM

Well:

Code:

$ ping -c3 compuserve.de
ping: unknown host compuserve.de
$ ping -c3 aol.com
PING aol.com (205.188.142.182) 56(84) bytes of data.

--- aol.com ping statistics ---
3 packets transmitted, 0 received, 100% packet loss, time 1999ms

compuserve.de does seem to be down, and aol.com is clearly just filtering and dropping icmp packets...

I think the majority of your problems is administrators dropping icmp packets (I do it). I do not think you have a problem with your setup. Every once in a while I have difficulties with the gentoo mirrors, so I just pick one that is close geographically and seems to be consistent. Since then, I have not had troubles.

Is compuserve.de pop3 a free service? Because if not you will want to give them a call about why their site is down.

As for NTP, post some specifics and maybe we can sniff out the problem.

bitpicker · 01-04-2006, 12:55 AM

The strange thing about the mirrors is that if the first isn't available, the others I have in make.conf usually aren't either; but for the moment I'm willing to accept it as coincidence.

Compuserve.de is free, and I can't find anything about a problem on their site (which of course doesn't mean anything), but at least today the reaction is slightly different - instead of thunderbird giving me an error 'connection refused' immediately, it tries to connect until a timeout occurs. Maybe it's going to be ok by February again at that rate...

Anyway, ntp is really giving me troubles. I have emerged it and followed the instructions from the gentoo wiki at http://gentoo-wiki.com/NTP. My ntp-client.conf contains the following lines:

Code:

NTPCLIENT_CMD="ntpdate"
NTPCLIENT_OPTS=" -b -u de.pool.ntp.org"
NTPCLIENT_TIMEOUT=30

According to the wiki I'm supposed to do the following next:

Code:

/etc/init.d/ntp-client start

the output of which (after timeout) is

Code:

 * Setting clock via the NTP client 'ntpdate' ...
/etc/init.d/ntp-client: line 34:  5959 Getötet                ${NTPCLIENT_CMD} ${NTPCLIENT_OPTS} >/dev/null
 * Failed to set clock                                                    [ !! ]

'Getötet' meaning 'killed'.

Using ntpdate as a command with the same parameters I get the following response after timeout:

Code:

ntpdate -b -u de.pool.ntp.org
 4 Jan 07:44:27 ntpdate[5984]: poll(): nfound = 0, error: Address family not supported by protocol

Adding the -d option I get slightly more:

Code:

 ntpdate -b -u -d de.pool.ntp.org
 4 Jan 07:46:25 ntpdate[5991]: ntpdate 4.2.0a@1.1190-r Sa Dez 31 16:37:04 CET 2005 (1)
 4 Jan 07:47:25 ntpdate[5991]: poll(): nfound = 0, error: Address family not supported by protocol

I have also used IP addresses found by netselect -s 3 de.pool.ntp.org instead of the pool itself, with the same results. Using the -4 option the results are the same, using -6 fails immediately with

Code:

ntpdate -b -u -d -6 de.pool.ntp.org
 4 Jan 07:51:20 ntpdate[6018]: ntpdate 4.2.0a@1.1190-r Sa Dez 31 16:37:04 CET 2005 (1)
Error : No address associated with hostname
 4 Jan 07:51:20 ntpdate[6018]: can't find host de.pool.ntp.org

 4 Jan 07:51:20 ntpdate[6018]: no servers can be used, exiting

I'm behind a DSL router with port-forwarding which I can't switch off; I have forwarded both UDP and TCP at port 123 to my machine, though with the -u option that isn't even necessary as I've been told. I don't want to set up an ntp server anyway, I just want my clock to tell me the real time...

Robin