Re: some remarks about fai 2.6, especialy on powerpc
> On Sun, 8 Aug 2004 12:45:32 -0400 (EDT), Paul Nijjar <[EMAIL PROTECTED]> said: > I personally prefer /usr/local to /var/lib, but I don't see it as > a serious bug because this is created by the sysadmin. (Would /var/local/ > be a bad place?) Me too. >> make-fai-nfsroot fails/stops after: >> Creating base.tgz >> mkdir: cannot create directory /var/lib/fai/nfsroot/var/state': File exists Never saw this error. > ! [ -d $NFSROOT/var/state ] || mkdir $NFSROOT/var/state Patch applied in 2.6.1. > One complaint from me: I want the "make-fai-nfsroot finished > $merror" line to be the final thing printed. I got confused and did not OK, I will change this behaviour. -- regards Thomas
Re: some remarks about fai 2.6, especialy on powerpc
On Sun, 8 Aug 2004, Holger Levsen wrote: > The default values of FAI_LOCATION (/usr/local/share/fai) and NFSROOT > (/usr/lib/fai/nfsroot) violate FHS 2.2. Should be /var/lib/fai/config and > /var/lib/fai/nfsroot, where I would even prefer /var/lib/fai/nfsroot/$ARCH > Thomas, we've already talked about this and FHS 2.3 (which is current, but not > used in sarge) says, FAI should use the new /srv-hierarchy instead, so you > said, you don't want to change the paths to not confuse people who are used > to FAI twice. I really think /usr/local/share/fai is a serious bug against > debian policies - but if noone else complains I won't submit it as a serious > bug. I personally prefer /usr/local to /var/lib, but I don't see it as a serious bug because this is created by the sysadmin. (Would /var/local/ be a bad place?) > make-fai-nfsroot fails/stops after: > Creating base.tgz > mkdir: cannot create directory /var/lib/fai/nfsroot/var/state': File exists > No diversion Diversion of /sbin/discover-modprobe by fai', none removed > > Unfortunatly I don't have the time _now_ (maybe tonite, hopefully tomorrow) to > look further into it, why this happens. make-fai-nfsroot on i386 worked just > fine here. On i386 I got the same thing this weekend. I did not really diagnose the problem (i.e. whether there is some logic error) but I patched the problem in the obvious way: *** make-fai-nfsrootSun Aug 8 12:19:27 2004 --- make-fai-nfsroot.newSun Aug 8 12:15:07 2004 *** *** 255,261 > lib/modules/$KERNELVERSION/modules.dep # dirty trick to hoax lvm echo 'NTPSERVERS=""' > etc/default/ntp-servers ! [ -d $NFSROOT/var/state ] || mkdir $NFSROOT/var/state [ "$verbose" ] && echo "Try to copy $cfdir/sources.list or /etc/apt/sources.list." cp -v $cfdir/sources.list $NFSROOT/etc/apt/sources.list || \ cp -v /etc/apt/sources.list $NFSROOT/etc/apt/sources.list || \ --- 255,261 > lib/modules/$KERNELVERSION/modules.dep # dirty trick to hoax lvm echo 'NTPSERVERS=""' > etc/default/ntp-servers ! mkdir $NFSROOT/var/state [ "$verbose" ] && echo "Try to copy $cfdir/sources.list or /etc/apt/sources.list." cp -v $cfdir/sources.list $NFSROOT/etc/apt/sources.list || \ cp -v /etc/apt/sources.list $NFSROOT/etc/apt/sources.list || \ == (I hope this patch is not in a bad format). One complaint from me: I want the "make-fai-nfsroot finished $merror" line to be the final thing printed. I got confused and did not realize that the nfsroot built properly because I saw some message about what was mounted when. (Yes, this was my fault but I see this as a usability flaw.) - Paul
some remarks about fai 2.6, especialy on powerpc
Hi, I just installed fai-2.6 in a sarge chroot on two woody machines (one i386, one powerpc), here are my remarks: The fai package recommends: syslinux and suggests: mknbi - both are i386 only. It would be nice (if possible) if these were arch-spezific. The default values of FAI_LOCATION (/usr/local/share/fai) and NFSROOT (/usr/lib/fai/nfsroot) violate FHS 2.2. Should be /var/lib/fai/config and /var/lib/fai/nfsroot, where I would even prefer /var/lib/fai/nfsroot/$ARCH Thomas, we've already talked about this and FHS 2.3 (which is current, but not used in sarge) says, FAI should use the new /srv-hierarchy instead, so you said, you don't want to change the paths to not confuse people who are used to FAI twice. I really think /usr/local/share/fai is a serious bug against debian policies - but if noone else complains I won't submit it as a serious bug. Line 36 in /etc/fai/make-fai-nfsroot.conf says "# following lines should be read only for you" - this is obviously (see below) only true if using i386. /usr/share/doc/fai/README.package_problems is mostly outdated and thus confusing / frightening, esp. after "The following table describes the problems with postinstall scripts." - I would suggest a rewrite. /usr/share/doc/links.html could include " * FAI on linux for powerpc: http://www.layer-acht.org/fai"; - I definitly update that page in future! and even before sarge is released ;-) Regarding the package fai-kernels I came up with the idea of making it arch-independed - with arch-specific kernel packages included. So we could start with three/four? packages inside (kernel-image.debs for i386, ppc, sparc and maybe amd64) - there is a working kernel-image.deb for ppc on layer-acht.org already, although at the moment it's to big to fit on a floppy. (fine for booting from net, hd or cd) packages.qa.debian.org/fai says "The package should be updated to follow the last version of Debian Policy (Standards-Version 3.6.1 instead of 3.5.6)." In /etc/fai/make-fai-nfsroot.conf for powerpc I had to prepend FAI_DEBOOTSTRAP_OPTS slightly differently: FAI_DEBOOTSTRAP_OPTS="--arch powerpc --exclude=grub,lilo,dmidecode,read-edid,hwtools,pcmcia-cs,ppp,pppconfig,pppoe,pppoeconf,dhcp-client,exim4,exim4-base,exim4-config,exim4-daemon-light,mailx,at,fdutils,info,modconf,libident,logrotate,exim" To packages for powerpc I had to add "quik yaboot" (powerpc bootloaders). make-fai-nfsroot fails/stops after: Creating base.tgz mkdir: cannot create directory /var/lib/fai/nfsroot/var/state': File exists No diversion Diversion of /sbin/discover-modprobe by fai', none removed Unfortunatly I don't have the time _now_ (maybe tonite, hopefully tomorrow) to look further into it, why this happens. make-fai-nfsroot on i386 worked just fine here. so far, so good ;-) regards, Holger P.S.: I'll keep an eye on these problems/issues and submit regular bug reports within the next days if neccessary.