Re: Problems with stability of networking / pfinet
Hi, On Tue, Dec 16, 2008 at 11:13:04PM +0100, Thomas Schwinge wrote: > On Sat, Jul 12, 2008 at 10:06:54AM +0200, I wrote: > > For a few weeks now, I seem to notice more and more problems w.r.t. > > networking. Olaf confirmed this. Well, not exactly: While I can confirm that disabling syslog fixed the ssh problems on some of Barry's boxes, the networking problems I'm experiencing on *my* machine seem to have a different nature: When apt-get downloads many packages during a larger upgrade, pfinet regularily hangs after a while. The current ssh connection consequently hangs as well. New connections fail, though IIRC in an unusual manner. (Don't remember the details.) After restarting pfinet and sshd, everything works again. (Well, I also need to create a new screen session -- the old session is not able to create new screens anymore, though the existing ones go on working perfectly fine...) I never tried disabling syslog on my box I think -- but I have a hard time believing this problem is related to syslog as well... -antrik-
Re: Problems with stability of networking / pfinet
Hello! Nothing fundamentally new; just noting down what we know so far. On Sat, Jul 12, 2008 at 10:06:54AM +0200, I wrote: > For a few weeks now, I seem to notice more and more problems w.r.t. > networking. Olaf confirmed this. > > $ ssh flubber > Warning: Permanently added '[flubber.bddebian.com]:2250' (RSA) to the > list of known hosts. > Connection closed by 71.224.175.179 > > The connection is first established, but then closed. > > $ ssh r...@flubber > Warning: Permanently added '[flubber.bddebian.com]:2250' (RSA) to the > list of known hosts. > r...@flubber.bddebian.com's password: > Connection closed by 71.224.175.179 > > The remote site is even asking for the password before then closing the > connection again. This difference is because the non-root case is using public key authentification. The underlying problem is that sshd hangs when it tries to syslog the new successful login attempt. (Samuel told about this already some time ago.) At this time everything using syslog will hang: be it a ``logger foo'' command manually run, or be it using sudo which roughly does the same logging sshd does. This situation can easily be induced like this: $ while logger .; do echo -n .; done This will usually hang after a few hundred invokations. Killing the system's pflocal server will ``fix'' this situation. If you, however, after that restart the system's syslogd, the looping command will again hang after another few hundred invokations. > A problem with the TCP/IP stack? But nothing has been changed > there, recently. Michael also assured me that he's still using GCC 4.2 > for building pfinet (using GCC 4.3 indeed yields a dysfunctional one). > Or a problem with the SSH packages? Killing pfinet and restarting sshd indeed also does ``fix'' this situation (why?), but neither pfinet nor sshd seem to be originally involved in the underlying problem. Regards, Thomas signature.asc Description: Digital signature
Re: Problems with stability of networking / pfinet
Hello! On Wed, Jul 16, 2008 at 10:45:38AM +0200, I wrote: > On Sat, Jul 12, 2008 at 10:06:54AM +0200, I wrote: > > $ ssh flubber > > Warning: Permanently added '[flubber.bddebian.com]:2250' (RSA) to the > > list of known hosts. > > Connection closed by 71.224.175.179 I logged in as `root' on gnubber, created an account `tschwinge', logged out, and tried to log in as `tschwinge' again, but: [EMAIL PROTECTED]'s password: Connection closed by 71.224.175.179 Logging in as `root' again fails the same way now. We really need to track this down. Regards, Thomas signature.asc Description: Digital signature
Re: Problems with stability of networking / pfinet
Hello! Barry rebooted flubber and I was able to do some (tiny amount of) work on her, but now... On Sat, Jul 12, 2008 at 10:06:54AM +0200, I wrote: > $ ssh flubber > Warning: Permanently added '[flubber.bddebian.com]:2250' (RSA) to the > list of known hosts. > Connection closed by 71.224.175.179 ... we're already again back to this one. I did not really stress her pfinet server instance, etc. No idea what's going on there. Goddamn. Regards, Thomas signature.asc Description: Digital signature
Re: Problems with stability of networking / pfinet
Hi, On Sat, Jul 12, 2008 at 10:06:54AM +0200, Thomas Schwinge wrote: > For a few weeks now, I seem to notice more and more problems w.r.t. > networking. > Any idea? A problem with the TCP/IP stack? But nothing has been > changed there, recently. Well, I started seeing it some two or three months ago, and it could have been in the archive much longer -- I don't upgrade very regularily... > Or a problem with the SSH packages? Seems unlikely: Even forcibly killing and restarting sshd doesn't restore it to sanity -- pfinet needs to be restarted fist, and only afterwards restarting sshd helps. In my case, the screen session also becomes partially disfunctional (when trying to create a new window it says that it can't find the shell...) -- I guess that's related to the fact that my home directory is on NFS, and thus also affected by pfinet going bonkers. The problem seems related to network load somehow: When doing a larger upgrade, it's sure to break while downloading the packages. (I seldom saw the problem occuring in other situations than apt downloading packages...) -antrik-
Problems with stability of networking / pfinet
Hello! For a few weeks now, I seem to notice more and more problems w.r.t. networking. Olaf confirmed this. $ ssh flubber Warning: Permanently added '[flubber.bddebian.com]:2250' (RSA) to the list of known hosts. Connection closed by 71.224.175.179 The connection is first established, but then closed. $ ssh [EMAIL PROTECTED] Warning: Permanently added '[flubber.bddebian.com]:2250' (RSA) to the list of known hosts. [EMAIL PROTECTED]'s password: Connection closed by 71.224.175.179 The remote site is even asking for the password before then closing the connection again. Barry did reboot the machine just yesterday, and it has been working just fine in the weeks/months before. Any idea? A problem with the TCP/IP stack? But nothing has been changed there, recently. Michael also assured me that he's still using GCC 4.2 for building pfinet (using GCC 4.3 indeed yields a dysfunctional one). Or a problem with the SSH packages? Regards, Thomas signature.asc Description: Digital signature
