Re: unpleasant ps output and possible related problems.
On Wed, Sep 05, 2001 at 11:21:06PM -0400, Mike Barcroft wrote: > I think it can safely be said that you're rebooting too much. The > process can be simplified to: > make world > make kernel > mergemaster > reboot For -current I would suggest a slight modification to this -- to make sure everything compiles before installing any bits: make buildworld make buildkernel make installkernel make installworld mergemaster reboot reboot between the installkernel and installworld steps if you like. -- -- David ([EMAIL PROTECTED]) To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: unpleasant ps output and possible related problems.
Dave Cornejo <[EMAIL PROTECTED]> writes: > you wrote: > > > When you rebuild and install a new kernel, are you also doing a > > `make buildworld` and a `make installworld` in /usr/src before you > > reboot? > > My usual method is to build a kernel, reboot, build world, reboot, > build a kernel using the new world, reboot again, do a mergemaster, > one final build world, reboot, then test. > > If I'm bored I'll do it all over again after combing for stale binaries > > Fortunately, I have a very fast system. :-) I think it can safely be said that you're rebooting too much. The process can be simplified to: make world make kernel mergemaster reboot Best regards, Mike Barcroft To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: unpleasant ps output and possible related problems.
you wrote: > When you rebuild and install a new kernel, are you also doing a > `make buildworld` and a `make installworld` in /usr/src before you > reboot? My usual method is to build a kernel, reboot, build world, reboot, build a kernel using the new world, reboot again, do a mergemaster, one final build world, reboot, then test. If I'm bored I'll do it all over again after combing for stale binaries Fortunately, I have a very fast system. :-) > Sometimes changes to userland are trivial, and you may not need to > rebuild userland, but utmp corruption is indicative of changes that > require userland be rebuilt and installed. > > Ideally, you should buildworld/installworld *EVERY* time you build a > -current kernel. > > Of course, if you have already done this, feel free to issue me a > boot to the head. I expect that this is the first question that should be asked of anyone who doesn't state explicitly that they follow a rigorous process for assuring a good build. No boot to the head necessary... > You note that you are running innd, please don't tell me that you >are using -current in a production environment... -current is always >subject to massive *FUNDAMENTAL* changes with only a moment's notice, >and breakage without any notice at all... Using -current in a >production environment, unless seriously justified [such as -current >being more stable than -stable], is a fine way to put yourself in a >position to commit hari-kari, and nobody wants that. I would call it a non-production system - besides, how better to test -current than by doing exactly what I would do with it in a real production system? I really don't think that this is an INN problem though - my best guess at the moment is that either something is busted in csh/tcsh or that something it relies upon is broken. The outward symptoms I saw that screwed up news or the boot sequence I think could be explained by the scripts returning control to console rather than exiting the shell, but that's a wild guess. I have rolled most of my -current systems back to a source tree from 23:36 UT on Monday night which is the last time I built a fully working system. I don't have too much time to play with it but I still have two very -current systems that exhibit the problem of the ps corruption so I'll keep plugging and if I get some time this weekend and they still are doing this, then I'll try and get more info. dave c -- Dave Cornejo @ Dogwood Media, Fremont, California (also [EMAIL PROTECTED]) "There aren't any monkeys chasing us..." - Xochi To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: unpleasant ps output and possible related problems.
Dave Cornejo wrote: > I apologize for not having any idea where to start on this. I am not > whining for someone to fix something, merely reporting an odd behavior > that I have now seen on multiple machines in cae it means something to > anybody. > > I am tracking current almost daily on three machines. Starting > yesterday I managed to get one box that refused to go into multiuser > mode it would run the rc script and then hang somewhere executing the > scripts in /usr/local/etc/rc.d. If I Ctrl-C'ed it - it would come up > in the single-user mode shell. no login prompt, just the shell. I > could however telnet into the thing most things seemed to work. > > In this state it had hung without starting INN - so I su'ed and tried > to start it. INN starts, but I end up at a prompt with a uid of news! > If I exit that, INN dies. > > I do a ps -ax and I get some corrupt lines: > > 471 p0 Is 0:00.07 -csh (csh) > 473 p0 I 0:00.01 su -m > 474 p0 S 0:00.04 \M-[\M-!\^D\b\M-X\M-!\^D\b (csh) > 12673 p0 R+ 0:00.00 ps ax > 466 v0 Is+0:00.01 /usr/libexec/getty Pc ttyv0 > > In troubleshooting this I went back to an older kernel and the problem > persists. Change back to an old world and it's gone. Tried the > new kernel with old world and it also seems to work fine. So the > problem seems to be somewhere in the libs or userland. > > Now I went and looked at some other systems rebuilt yesterday evening > and today and while they still work I see the same sort of corruption > as above in the ps output - but no other apparent side effects. > > The corrupted line shows up in many different places and users, and > the exact contents vary, but there's always a "(csh)" at the end. When you rebuild and install a new kernel, are you also doing a `make buildworld` and a `make installworld` in /usr/src before you reboot? Sometimes changes to userland are trivial, and you may not need to rebuild userland, but utmp corruption is indicative of changes that require userland be rebuilt and installed. Ideally, you should buildworld/installworld *EVERY* time you build a -current kernel. Of course, if you have already done this, feel free to issue me a boot to the head. You note that you are running innd, please don't tell me that you are using -current in a production environment... -current is always subject to massive *FUNDAMENTAL* changes with only a moment's notice, and breakage without any notice at all... Using -current in a production environment, unless seriously justified [such as -current being more stable than -stable], is a fine way to put yourself in a position to commit hari-kari, and nobody wants that. jim -- ET has one helluva sense of humor! He's always anal-probing right-wing schizos! POWER TO THE PEOPLE! _ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
unpleasant ps output and possible related problems.
I apologize for not having any idea where to start on this. I am not whining for someone to fix something, merely reporting an odd behavior that I have now seen on multiple machines in cae it means something to anybody. I am tracking current almost daily on three machines. Starting yesterday I managed to get one box that refused to go into multiuser mode it would run the rc script and then hang somewhere executing the scripts in /usr/local/etc/rc.d. If I Ctrl-C'ed it - it would come up in the single-user mode shell. no login prompt, just the shell. I could however telnet into the thing most things seemed to work. In this state it had hung without starting INN - so I su'ed and tried to start it. INN starts, but I end up at a prompt with a uid of news! If I exit that, INN dies. I do a ps -ax and I get some corrupt lines: 471 p0 Is 0:00.07 -csh (csh) 473 p0 I 0:00.01 su -m 474 p0 S 0:00.04 \M-[\M-!\^D\b\M-X\M-!\^D\b (csh) 12673 p0 R+ 0:00.00 ps ax 466 v0 Is+0:00.01 /usr/libexec/getty Pc ttyv0 In troubleshooting this I went back to an older kernel and the problem persists. Change back to an old world and it's gone. Tried the new kernel with old world and it also seems to work fine. So the problem seems to be somewhere in the libs or userland. Now I went and looked at some other systems rebuilt yesterday evening and today and while they still work I see the same sort of corruption as above in the ps output - but no other apparent side effects. The corrupted line shows up in many different places and users, and the exact contents vary, but there's always a "(csh)" at the end. -- Dave Cornejo @ Dogwood Media, Fremont, California (also [EMAIL PROTECTED]) "There aren't any monkeys chasing us..." - Xochi To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
