Reply on 20-12-2006 15:18:44 <<<
On Tue, Dec 19, 2006 at 08:53:13PM -0500, Patrick Shanahan wrote:
me thinks it not be a 10.2 problem, but a kernel problem. I have
10.1
with 2.6.18.5-jen40-default and cannot access the site. I have
read
the thread and, as I recall, noone with 2.6.18+ could access, but several with 2.6.16 could.
I am running 10.1 with a self-compiled "Linux seahunt 2.6.18"
kernel,
and it raises the Marymount site just fine. If it's a 2.6.18 problem, it's related to something in the SUSE version of 2.6.18, because kernel.org's
2.6.18
works fine.
Michael
Michael,
just to be sure, as there was some confusion: Are you trying www.marymount.edu ? THAT is the one in question. Or www.keh.com Please retest and repost.
Dominique
On Wed, Dec 20, 2006 at 02:19:34PM +0100, Dominique Leuenberger wrote:
I am running 10.1 with a self-compiled "Linux seahunt 2.6.18"
kernel,
and it raises the Marymount site just fine. If it's a 2.6.18 problem, it's related to something in the SUSE version of 2.6.18, because kernel.org's
2.6.18
works fine.
Yeah, I got snagged by the wrong address. www.marymount.com works, www.marymount.edu and www.keh.com don't. So just to be clear, 10.1 with kernel.org 2.6.18 does NOT work.
Michael
Dominique Leuenberger wrote:
just to be sure, as there was some confusion: Are you trying www.marymount.edu ? THAT is the one in question. Or www.keh.com Please retest and repost.
Sorry to confuse the issue, it was an honest typo. I was typing it from memory at the end of a long day (that just became a bit longer). Anyway, thanks Leendert for catching my mistake before it really muddled the problem too bad. I tried here at home on 10.2 to start squid to see if the cache was the reason the 10.2 at work connected (or maybe because the cache is on a 9.3 server). Even running through Squid here at home, it doesn't work. So it isn't just a proxy. But, I found this interesting info from squid in the log. Dec 20 21:41:58 jmorris squid[12862]: WARNING: Closing client 192.168.10.1 connection due to lifetime timeout Dec 20 21:41:58 jmorris squid[12862]: http://www.marymount.edu/ Dec 20 21:41:58 jmorris squid[12862]: WARNING: Closing client 192.168.10.1 connection due to lifetime timeout Dec 20 21:41:58 jmorris squid[12862]: http://www.marymount.edu/
I am not sure what that means exactly, but it still doesn't work here, even with a proxy. So at work it must have been because the server is still running 9.3.
On Wed, 2006-12-20 at 21:52 +0800, Joe Morris (NTM) wrote:
Dominique Leuenberger wrote:
just to be sure, as there was some confusion: Are you trying www.marymount.edu ? THAT is the one in question. Or www.keh.com Please retest and repost.
Sorry to confuse the issue, it was an honest typo. I was typing it from memory at the end of a long day (that just became a bit longer). Anyway, thanks Leendert for catching my mistake before it really muddled the problem too bad. I tried here at home on 10.2 to start squid to see if the cache was the reason the 10.2 at work connected (or maybe because the cache is on a 9.3 server). Even running through Squid here at home, it doesn't work. So it isn't just a proxy. But, I found this interesting info from squid in the log. Dec 20 21:41:58 jmorris squid[12862]: WARNING: Closing client 192.168.10.1 connection due to lifetime timeout Dec 20 21:41:58 jmorris squid[12862]: http://www.marymount.edu/ Dec 20 21:41:58 jmorris squid[12862]: WARNING: Closing client 192.168.10.1 connection due to lifetime timeout Dec 20 21:41:58 jmorris squid[12862]: http://www.marymount.edu/
I am not sure what that means exactly, but it still doesn't work here, even with a proxy. So at work it must have been because the server is still running 9.3.
If the proxy were running on anything but 10.2 w/ 2.6.18 kernel it would probably work.
Kenneth Schneider wrote:
If the proxy were running on anything but 10.2 w/ 2.6.18 kernel it would probably work.
So it does seem to be a definite kernel bug. weird.
On 2006-12-20 16:21, Charles philip Chan wrote:
On 20 Dec 2006, Joe_Morris@ntm.org wrote:
So it does seem to be a definite kernel bug. weird.
They don't work with kernel 2.6.19 either.
Charles
Incorrect. Felix Miata has compiled one which works, with a lot of stuff stripped out:
On 2006/12/20 09:01 (GMT+-0500) Nick Zentena apparently typed:
On Wednesday 20 December 2006 09:02, Felix Miata wrote:
Have you read the bug? Etch's and Fedora's 2.6.18 kernels also fail. OTOH, the vanilla 2.6.19 kernel worked for me yesterday, and still works today.
How did you configure it? I downloaded it and did an make oldconfig. It still fails here.
I didn't build it specifically for testing this problem, instead for cifs & smbfs testing. I stripped out a lot of useless hardware support to conserve space. http://mrmazda.no-ip.com/tmp/config
Maybe we can talk Felix into adding stuff back in, until it breaks -- then we will have found the culprit :-)
On Thursday 21 December 2006 00:42, Darryl Gregorash wrote:
On 2006-12-20 16:21, Charles philip Chan wrote:
On 20 Dec 2006, Joe_Morris@ntm.org wrote:
So it does seem to be a definite kernel bug. weird.
They don't work with kernel 2.6.19 either.
Charles
Incorrect. Felix Miata has compiled one which works, with a lot of stuff
stripped out:
On 2006/12/20 09:01 (GMT+-0500) Nick Zentena apparently typed:
On Wednesday 20 December 2006 09:02, Felix Miata wrote:
Have you read the bug? Etch's and Fedora's 2.6.18 kernels also fail. OTOH, the vanilla 2.6.19 kernel worked for me yesterday, and still works today.
How did you configure it? I downloaded it and did an make oldconfig. It still fails here.
I didn't build it specifically for testing this problem, instead for cifs & smbfs testing. I stripped out a lot of useless hardware support to conserve space. http://mrmazda.no-ip.com/tmp/config
Maybe we can talk Felix into adding stuff back in, until it breaks -- then we will have found the culprit :-)
Adding back in: not 1 by 1, but 50% by 50%. With 16 items, you would need 4 runs (2^4=16)..., maybe less if you leave unrelated stuff out. ;-P
Cheers,
Leen
On 20 Dec 2006, raven@accesscomm.ca wrote:
Incorrect. Felix Miata has compiled one which works, with a lot of stuff stripped out:
On 2006/12/20 09:01 (GMT+-0500) Nick Zentena apparently typed:
On Wednesday 20 December 2006 09:02, Felix Miata wrote:
Have you read the bug? Etch's and Fedora's 2.6.18 kernels also fail. OTOH, the vanilla 2.6.19 kernel worked for me yesterday, and still works today.
Strange, my 2.6.19 is a vanilla kernel with the TIOCGDEV patch and the CK patchset:
http://members.optusnet.com.au/ckolivas/kernel/
For the record I have no problems access other websites. The only ones that I have problems with are the 2 sites that the OP mentioned.
Charles
On 12/20/06, Joe Morris (NTM) Joe_Morris@ntm.org wrote:
Kenneth Schneider wrote:
If the proxy were running on anything but 10.2 w/ 2.6.18 kernel it would probably work.
So it does seem to be a definite kernel bug. weird.
Once again, my configuration at home is: Pentium III box with openSuSE 10.2, default kernel 2.6.18 and 2 NICs. One NIC is connected to ADSL modem. The box runs SuSE Firewall2 with masquerading. I can access both sites FROM THIS BOX without any problem.
I also tried marymount site before I started firewall, that is it was only PPPoE (for ADSL modem), no NAT and no proxy. It was OK as well.
Well, of course there was my provider ...
On Wednesday 20 December 2006 22:09, Mark Goldstein wrote:
Well, of course there was my provider ...
Since your results differ from most with the same kernel, have you tested to see if you are behind a transparent proxy at your provider?
On 12/21/06, John Andersen jsa@pen.homeip.net wrote:
On Wednesday 20 December 2006 22:09, Mark Goldstein wrote:
Well, of course there was my provider ...
Since your results differ from most with the same kernel, have you tested to see if you are behind a transparent proxy at your provider?
No I did not. Is there any tool that can assist in that? Of course I can use tracerout, but how would I know whether any of intermediate hosts was proxy?
On Wednesday 20 December 2006 22:32, Mark Goldstein wrote:
On 12/21/06, John Andersen jsa@pen.homeip.net wrote:
On Wednesday 20 December 2006 22:09, Mark Goldstein wrote:
Well, of course there was my provider ...
Since your results differ from most with the same kernel, have you tested to see if you are behind a transparent proxy at your provider?
No I did not. Is there any tool that can assist in that? Of course I can use tracerout, but how would I know whether any of intermediate hosts was proxy?
-- Mark Goldstein
This page has some info that might help: http://tracetcp.sourceforge.net/usage_proxy.html but it might require you install that package.
Also the tcptraceroute package can help: (slaged this off a google search)
Find it at: http://michael.toren.net/code/tcptraceroute/
tcptraceroute servername
will do a TCP traceroute on port 80.
tcptraceroute servername 25
will do a TCP traceroute on port 25 to tell you what hops your TCP packets take to get to the host. Because the ICMP route, and the TCP route might be a bit different because of router configs.
tcptraceroute www.slashdot.org
1 * * * 2 slashdot.org (66.35.250.150) [open] 0.983 ms 0.838 ms 0.347 ms
Hmmm, only two hops from me to slashdot? not right, I should at least see the IP's to get to my upstream provider.... Proxy server before I even get to my gateway.
On 12/21/06, John Andersen jsa@pen.homeip.net wrote:
This page has some info that might help: http://tracetcp.sourceforge.net/usage_proxy.html but it might require you install that package.
Also the tcptraceroute package can help: (slaged this off a google search)
Find it at: http://michael.toren.net/code/tcptraceroute/
tcptraceroute servername
will do a TCP traceroute on port 80.
tcptraceroute servername 25
will do a TCP traceroute on port 25 to tell you what hops your TCP packets take to get to the host. Because the ICMP route, and the TCP route might be a bit different because of router configs.
tcptraceroute www.slashdot.org
1 * * * 2 slashdot.org (66.35.250.150) [open] 0.983 ms 0.838 ms 0.347 ms
Hmmm, only two hops from me to slashdot? not right, I should at least see the IP's to get to my upstream provider.... Proxy server before I even get to my gateway.
John, Thank you for information. You are absolutely right. tcptraceroute ends in 2 hops, while traceroute makes 20+! So my yes, my provider uses transparent proxy. I'll add this information to bugzilla
On 2006-12-21 04:44, John Andersen wrote:
<snip> Also the tcptraceroute package can help: (slaged this off a google search)
Find it at: http://michael.toren.net/code/tcptraceroute/
Also available on Pascal Bleser's repository, ftp://ftp.gwdg.de/pub/linux/misc/suser-guru/rpm/<version>/
Darryl,
On Thursday 21 December 2006 13:51, Darryl Gregorash wrote:
...
Find it at: http://michael.toren.net/code/tcptraceroute/
The latest source (1.5beta7) builds on 10.2 and 10.0, if you install all the required development packages.
Also available on Pascal Bleser's repository, ftp://ftp.gwdg.de/pub/linux/misc/suser-guru/rpm/<version>/
There appears to be a tcptraceroute package in the Guru's RPM repository for 10.0 but not for 10.2. At least not at this moment.
Randall Schulz
On 2006-12-21 16:31, Randall R Schulz wrote:
There appears to be a tcptraceroute package in the Guru's RPM repository for 10.0 but not for 10.2. At least not at this moment.
I'm still running in the Dark Ages, I only checked 9.3 :-)
All the builds are from August, and there is only one src.rpm, supposedly for 10.0. Anyone in dire need of an rpm package for 10.2 could try that:
ftp://ftp.gwdg.de/pub/linux/misc/suser-guru/rpm/packages/Network/tcptraceroute/src/tcptraceroute-1.5beta7-1.guru.suse100.src.rpm
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Darryl Gregorash wrote:
On 2006-12-21 16:31, Randall R Schulz wrote:
There appears to be a tcptraceroute package in the Guru's RPM repository for 10.0 but not for 10.2. At least not at this moment.
I'm still running in the Dark Ages, I only checked 9.3 :-) All the builds are from August, and there is only one src.rpm, supposedly for 10.0. Anyone in dire need of an rpm package for 10.2
The src.rpm is rebuildable on all SUSE versions. It's just that I cannot suppress that ".suse???." in the src.rpm while having it in the binary RPMs ;)
could try that: ftp://ftp.gwdg.de/pub/linux/misc/suser-guru/rpm/packages/Network/tcptraceroute/src/tcptraceroute-1.5beta7-1.guru.suse100.src.rpm
10.2 builds will be available in my repository in a few hours, after the next sync with the server + mirrors.
For the impatient, it rebuilds just fine on 10.2: rpmbuild --rebuild --target=i586 \ tcptraceroute-1.5beta7-1.guru.suse100.src.rpm
cheers - -- -o) Pascal Bleser http://linux01.gwdg.de/~pbleser/ /\ pascal.bleser@skynet.be guru@unixtech.be __v The more things change, the more they stay insane.
-
Charles philip Chan -
Darryl Gregorash -
Dominique Leuenberger -
Joe Morris (NTM) -
John Andersen -
Kenneth Schneider -
Leendert Meyer -
Mark Goldstein -
Michael Nelson -
Pascal Bleser -
Randall R Schulz