Re: Is anyone using woody in a production environment?
From: Angus D Madden [EMAIL PROTECTED] Reply-To: [EMAIL PROTECTED] To: Roderick Cummings [EMAIL PROTECTED] CC: debian-user@lists.debian.org Subject: Re: Is anyone using woody in a production environment? Date: Fri, 11 Jan 2002 09:28:46 -0500 On Fri, Jan 11, 2002 at 02:32:49AM -0500, Roderick Cummings wrote: Only that the new 2.4 kernel-image wont mount my root partition: request_module[block-major-3]: Root fs not mounted VFS: Cannot open root device 303 or 03:03 Please append a correct root= boot option Kernenl Panic: VFS Unable to mount root fs on 03:03 Even though lilo.conf specifies the correct root= option, and fstab is correct as well. I experienced the same problem repeatedly on every box I tried to install the 2.4 kernel-image debs. Other people I talked to didn't have the same problem. It's because the new kernels are initrd. Rather than read up on mkinitrd, I think it's easier to just download the kernel source and compile your own kernel. But that's just me. g attach3 hmmm, I wonder what we could be doing wrong... _ Send and receive Hotmail on your mobile device: http://mobile.msn.com
Re: Is anyone using woody in a production environment?
Is there a safe and stable way to build/install woody packages onto a potato system other than to dist-upgrade to woody? what's from with dist-upgrade? Only that the new 2.4 kernel-image wont mount my root partition: request_module[block-major-3]: Root fs not mounted VFS: Cannot open root device 303 or 03:03 Please append a correct root= boot option Kernenl Panic: VFS Unable to mount root fs on 03:03 Even though lilo.conf specifies the correct root= option, and fstab is correct as well. _ Join the worlds largest e-mail service with MSN Hotmail. http://www.hotmail.com
Re: Is anyone using woody in a production environment?
On Fri, Jan 11, 2002 at 02:32:49AM -0500, Roderick Cummings wrote: Only that the new 2.4 kernel-image wont mount my root partition: request_module[block-major-3]: Root fs not mounted VFS: Cannot open root device 303 or 03:03 Please append a correct root= boot option Kernenl Panic: VFS Unable to mount root fs on 03:03 Even though lilo.conf specifies the correct root= option, and fstab is correct as well. I experienced the same problem repeatedly on every box I tried to install the 2.4 kernel-image debs. Other people I talked to didn't have the same problem. It's because the new kernels are initrd. Rather than read up on mkinitrd, I think it's easier to just download the kernel source and compile your own kernel. But that's just me. g pgpu5JzxPqcED.pgp Description: PGP signature
Re: Is anyone using woody in a production environment?
On Fri, Jan 11, 2002, Roderick Cummings wrote: Is there a safe and stable way to build/install woody packages onto a potato system other than to dist-upgrade to woody? what's from with dist-upgrade? Only that the new 2.4 kernel-image wont mount my root partition: request_module[block-major-3]: Root fs not mounted VFS: Cannot open root device 303 or 03:03 Please append a correct root= boot option Kernenl Panic: VFS Unable to mount root fs on 03:03 Even though lilo.conf specifies the correct root= option, and fstab is correct as well. Did you remember to add the initrd line to lilo.conf? HTH, Daniel -- Daniel A. Freedman Laboratory for Atomic and Solid State Physics Department of Physics Cornell University
Re: Is anyone using woody in a production environment?
On Fri, Jan 11, 2002 at 02:32:49AM -0500, Roderick Cummings wrote: | | | Is there a safe and stable way to build/install woody packages onto a | potato system other than to dist-upgrade to woody? | | what's from with dist-upgrade? | | Only that the new 2.4 kernel-image wont mount my root partition: | | request_module[block-major-3]: Root fs not mounted | VFS: Cannot open root device 303 or 03:03 | Please append a correct root= boot option | Kernenl Panic: VFS Unable to mount root fs on 03:03 | | Even though lilo.conf specifies the correct root= option, and fstab is | correct as well. Did you load the initrd so that the kernel can get the driver for your disk controller and root filesystem? -D -- In my Father's house are many rooms; if it were not so, I would have told you. I am going there to prepare a place for you. And if I go and prepare a place for you, I will come and take you to be with me that you also may be where I am. John 14:2-3
RE: Is anyone using woody in a production environment?
I am using woody as a postgresql server in production. I wanted version 7 and testing is *mostly* safe if you don't go hog wild installing packages. I haven't had too many problems. | 2) We upgrade to testing. | | Is it safe? image of Marathon Man. Who is running production servers | on testing? what if any issues have arisen? I'm actually running 2 out of 6 boxes on testing (4 production, 2 playground). Other than the expected X configuration problems and a few ill-timed keystrokes, I am doing fine. I have not gone to a 2.4 kernel in production but I am comfortably running 2.2.20. For the production machine I started with a clean install of potato and went straight to woody. Less problems than a configured system upgrade. | 3) We build the Debian packages from testing on stable. IMO not a good idea... dependencies, dependencies, dependencies. You'll end up basically upgrading to woody anyway, just do it to start with. Hope this helps, Brooks Brooks R. Robinson ERP Systems Administrator/Accounting Manager Chance Industries, Inc. 4219 Irving Wichita, Kansas 67207 Work: (316)945-6555x2632 Email: [EMAIL PROTECTED]
Re: Is anyone using woody in a production environment?
john wrote: Hello everyone! We have come to a point where we would like to use some software that is currently in woody on a production server that is currently running potato. Like what software? Maybe some Debian maintainer has made some unofficial debs for these software? Make some effort to find these packages. Google is your friend. However, when you use other packages other than those sanctioned by Debian, you might compromise your system. Now we have a few approaches as I see it: 1) We install the packages from upstream source into /usr/local/ OK, we're talking about openssl, curl-ssl, perl, libssl, apache, apache-ssl, php etc. Might be a nightmare to maintain in the future. And since we'll be selling boxes with all this set up on them it's going to make things much more expensive for us. I got a copy of the unofficial CDs from fsn.hu. They have some fairly recent versions of these ssl-enabled software that were backported from Sid to Potato. I suggest you try get these CD's for they might have what you are looking for. 2) We upgrade to testing. Is it safe? image of Marathon Man. Who is running production servers on testing? what if any issues have arisen? I've even got a friend who runs Sid on a production environment with a 2.4 kernel. He manages to get it through very smoothly, and I really wonder how he tamed that crazy distro w/c is even wilder than woody. But he still maintains a Potato with 2.2 kernel as the main server. But still, for production environment, I recommend to stick with Potato for absolute stability (well, almost... but at least you get less worries) 3) We build the Debian packages from testing on stable. I've tried this, and either got it wrong quite likely or it just doesn't work like that as build curl-ssl then wants perl, which doesnt want libdbi-perl. It wants a libc6 upgrade. Which might (will it?) break other things etc etc. Any advice as to how to best manage this is appreciated - I'm particularly interested in the opinions of anyone who has actually _done_ this sort of thing. Is there a safe and stable way to build/install woody packages onto a potato system other than to dist-upgrade to woody? There is, but it takes lots of tweaking of file dependencies. AS IN LOTS. But with all the effort, it's not really a good road ahead. I know for I did this a lot in the past. But not on production environments - I only do that on my Potato box at home. Tweaking the file dependencies was the only way to make my own custom Debian packages for Mozilla 0.9.7, Galeon 1.0, XMovie et.al to work with Potato. So to avoid too much hardships, try looking for the unofficial packages or try the fsn.hu unofficial CD's et.al. Paolo Falcone __ www.edsamail.com
Re: Is anyone using woody in a production environment?
john [EMAIL PROTECTED] writes: 2) We upgrade to testing. Is it safe? image of Marathon Man. Who is running production servers on testing? what if any issues have arisen? We run our production servers on testing. I haven't run into any problems. We do an upgrade approximately once a week, but I watch for security issues with packages and upgrade those when I see them. We upgrade the dev boxes two days before the production boxes, and any packages on the production boxes that I know we use regularly I put on hold so that we get at least 2+n days of testing of that package on the dev boxes (n = the number of days I keep the package on hold, and is quite large, 30 for some packages). So far, we haven't seen any problems. 3) We build the Debian packages from testing on stable. I've tried this, and either got it wrong quite likely or it just doesn't work like that as build curl-ssl then wants perl, which doesnt want libdbi-perl. It wants a libc6 upgrade. Which might (will it?) break other things etc etc. Yes, this is the approach we used to use, until the mess of dependencies got so bad that we just bit the bullet and upgraded to woody. Once the new perl and libc made it into testing, it got much harder to build and install woody packages on potato. -- Dave Carrigan ([EMAIL PROTECTED])| Yow! You mean now I can SHOOT UNIX-Apache-Perl-Linux-Firewalls-LDAP-C-DNS | YOU in the back and further BLUR Seattle, WA, USA| th' distinction between FANTASY http://www.rudedog.org/ | and REALITY?
Re: Is anyone using woody in a production environment?
also sprach Brooks R. Robinson [EMAIL PROTECTED] [2002.01.11.1700 +0100]: | 3) We build the Debian packages from testing on stable. IMO not a good idea... dependencies, dependencies, dependencies. You'll end up basically upgrading to woody anyway, just do it to start with. disagreed. do you have an example? for instance, back in the days when i still deployed potato, i installed OpenSSH 2.9pl2 from a custom package. sure, it needed a newer openssl and a newer libssl0.9.6, but that was it. you don't end up upgrading to woody... -- martin; (greetings from the heart of the sun.) \ echo mailto: !#^.*|tr * mailto:; [EMAIL PROTECTED] la lune, c'est comme les canards il faut aimer caresser les chats pour avoir envie d'y aller. pgpQHiHR9tdVo.pgp Description: PGP signature
Is anyone using woody in a production environment?
Hello everyone! We have come to a point where we would like to use some software that is currently in woody on a production server that is currently running potato. Now we have a few approaches as I see it: 1) We install the packages from upstream source into /usr/local/ OK, we're talking about openssl, curl-ssl, perl, libssl, apache, apache-ssl, php etc. Might be a nightmare to maintain in the future. And since we'll be selling boxes with all this set up on them it's going to make things much more expensive for us. 2) We upgrade to testing. Is it safe? image of Marathon Man. Who is running production servers on testing? what if any issues have arisen? 3) We build the Debian packages from testing on stable. I've tried this, and either got it wrong quite likely or it just doesn't work like that as build curl-ssl then wants perl, which doesnt want libdbi-perl. It wants a libc6 upgrade. Which might (will it?) break other things etc etc. Any advice as to how to best manage this is appreciated - I'm particularly interested in the opinions of anyone who has actually _done_ this sort of thing. Is there a safe and stable way to build/install woody packages onto a potato system other than to dist-upgrade to woody? Thanks in Advance, John P Foster http://www.golden-orb.com
Re: Is anyone using woody in a production environment?
On Fri, Jan 11, 2002 at 12:42:20PM +1000, john wrote: :2) We upgrade to testing. : :Is it safe? image of Marathon Man. Who is running production servers :on testing? what if any issues have arisen? Well, not production servers. I do have a mix of workstations some testing, some stable, and a few I've pulled tricks on. We don't support testing so the fact that I haven't heard complaints doesn't indicate a lack of problems. But in a few cases where someone had a demonstrated need for newer tools than stable provides (wxpython, XFree4) I've switched to testing installed the required packages (which ofcourse pulls the new glibc) and then switched sources.list back to stable. This has worked extremely well. As I said these are user workstations not mission critical. If you can install a test system and then stress test it, then you'll see if this can work for you.
Re: Is anyone using woody in a production environment?
also sprach john [EMAIL PROTECTED] [2002.01.11.0342 +0100]: We have come to a point where we would like to use some software that is currently in woody on a production server that is currently running potato. it's about time ;^ Now we have a few approaches as I see it: 1) We install the packages from upstream source into /usr/local/ OK, we're talking about openssl, curl-ssl, perl, libssl, apache, apache-ssl, php etc. Might be a nightmare to maintain in the future. And since we'll be selling boxes with all this set up on them it's going to make things much more expensive for us. yes, it'll be a nightmare. i usually do this for 1-4 packages at the most, if debian hasn't packaged them. but i *hate* it. i am thoroughly spoiled by Debian ;) 2) We upgrade to testing. Is it safe? image of Marathon Man. Who is running production servers on testing? what if any issues have arisen? we are. at least 15 of them. i can give you exact counts tomorrow. sure it's safe... noone forces you to apt-get dist-upgrade everyday (there are bugs that come in that way), and you'll have to get down with apt's pinning to make sure to keep the versions you want and not to upgrade them. furthermore, you have to keep track of what's secure and what isn't. (we're a network security company, so that's a nice side-effect). 3) We build the Debian packages from testing on stable. I've tried this, and either got it wrong quite likely or it just doesn't work like that as build curl-ssl then wants perl, which doesnt want libdbi-perl. It wants a libc6 upgrade. Which might (will it?) break other things etc etc. you need a lot of -dev packages for this to work. my advice (i do this occassionally to get packages from unstable into testing): get yourself a $500 pizzabox machine that you'll use for exclusive repackaging, install basically all *-dev on it, along with debhelper et al., and then it should work. after all, the maintainers do nothing else (okay, they don't have all *-dev (i do), but you also don't need all. just try dpkg-buildpackage until it works, fixing dependencies as you go. or check Build-Depends of the package...) Any advice as to how to best manage this is appreciated - I'm particularly interested in the opinions of anyone who has actually _done_ this sort of thing. i do all three. we're running woody, have been for half a year (it's getting better by the day, one can *feel* the impending freeze), i occassionally get packages from unstable into the woody systems (early) by method 3, and some software that's not available as .deb, i install from tarball in /usr/local. if i have time, i try to package it into a .deb just to make dpkg know about it. Is there a safe and stable way to build/install woody packages onto a potato system other than to dist-upgrade to woody? what's from with dist-upgrade? but you can just update your sources.list to woody, then simply install what you want, and watch as *a lot* of dependencies stroll along... libc... base stuff... a bunch. -- martin; (greetings from the heart of the sun.) \ echo mailto: !#^.*|tr * mailto:; [EMAIL PROTECTED] the remote desktop feature of windows xp is really nice (and *novel*!). as a micro$oft consultant can *remotely* disable the personal firewall and control the system. we'll ignore the fact that this tampering with the firewall is not logged, and more importantly, that the firewall isn't restored when the clowns from redmod are done with their job. pgpRaoq0pmLQd.pgp Description: PGP signature
Re: Is anyone using woody in a production environment?
On Fri, Jan 11, 2002 at 12:42:20PM +1000, john wrote: 2) We upgrade to testing. Is it safe? image of Marathon Man. Who is running production servers on testing? what if any issues have arisen? Depends what you are running in what environment. RedHat is usually less stable then testing so I don't know what to tell you..
Re: Is anyone using woody in a production environment?
I agree with Martin ... you can smell a freeze coming. It's fantastic. I use woody for two Compaq production intranet servers, and a mix of potato/woody for one important server that I cannot risk something going wrong on during an upgrade. All three servers are about 450km from me, hence the caution. For a pair of Australian outback internet cafes, I'm using potato in one because I've not needed any new packages, and woody in the other because I did. A set of four training workstations I've pushed to sid to get galeon. For my home telecommute network, my gateway is potato with source rebuilds of one or two woody packages, and the eight other machines are woody. For four computers that I've donated to farmers or ranchers in the area, they are on woody just to get the latest tested desktop features; gnome, abiword, gnumeric, gnucash ... and for some of them I've pushed them to sid just to get galeon and evolution going. Most people I know on IRC with Debian production servers are using woody. They need too many of the recent features to justify potato. -- James Cameron ([EMAIL PROTECTED]) http://quozl.linux.org.au/ (or) http://quozl.netrek.org/
Re: Is anyone using woody in a production environment?
On Fri, Jan 11, 2002 at 12:42:20PM +1000, john wrote: We have come to a point where we would like to use some software that is currently in woody on a production server that is currently running potato. Now we have a few approaches as I see it: 1) We install the packages from upstream source into /usr/local/ I only do this when the debs aren't available or they don't work together nicely (eg vpopmail + courier-imap). Also when I need a compile option that isn't in the deb. Last resort. 2) We upgrade to testing. We have at least 10 machines running testing in a production environment. Provided you handle things carefully, it is definitely workable. Here's what I recommend: 0. Be conservative. Only upgrade when you absolutely positively have to. 1. Do all your upgrades on a test box first. Very important. If you run them blindly on a production server it feels like diving off a 10 meter platform - you have a moment of concious fear before you hit the water. 2. Keep everything in your /var/cache/apt/archives/. Being able to rollback to an earlier version is important. 3. If you are upgrading or installign a single package and apt gives you a long laundry list of packages. Think real long and hard before choosing Y. In my experience, [ probability of meltdown ] ~= [ number of packages to be upgraded ]^2 4. When you casually ignore recommendations 1-4, and the machine gets into an unworkable state (like the time when dpkg erased the link at /usr/bin/perl) don't panic. Go through your usual troublshooting routine (google, lists.debian.org, man, HOWTO, mailing list, irc, repeat). The answer will come. Hopefully it will come before people start to wake up and log into the machine. 5. Test everything throughly after the upgrade is completed - there may be subtle problems which only affect your custom application of the server and don't show up on the test box. 3) We build the Debian packages from testing on stable. I think this is definitely workable if you have an in-depth knowledge of how debs are made. If you have the time and the knowledge, it might be the best solution. You could also look for unofficial debs which have already been put together by other debian users. my $0.02 g pgpkJrYRIyw6s.pgp Description: PGP signature