Well, my prevoius post discussd difficulties until this point. But this is the end of the line. Rolling back to FC6.
Whats happened since the last post ?
Still playing with attempt 5.
Just glad I made a symbolic link for /var/cache/yum to one of my other filesystems
Removed all the packages that were failing dependencies, and once more ‘yum –obsoletes –disablerepo=*update* localupdate *’
Ok I must not understand localupdate, it seemed to check the rpm files ?, but then downloaded and installed over 2Gb from the net.
But still not there, the kernel is still fc6. Tried a ‘yum –disablereporepo=*update* upgrade kernel’, wouldn’t you know it, a missing dependency on ‘hal’. I think I need hal, not removing that yet.
So in frustration another ‘yum –disablerepo=*update* –obsoletes localupdate *’, bloody hell, it found another 1.5Gb to download off the net; but no missing dependency this time so it probably won’t get the kernel ???.
After an hout of download time
warning: rpmts_HdrFromFdno: Header V3 DSA signature: NOKEY, key ID 4ebfc273
GPG key retrieval failed: [Errno 5] OSError: [Errno 2] No such file or directory: ‘/etc/pki/rpm-gpg/RPM-GPG-KEY-fedora-i386’
You have new mail in /var/spool/mail/root
[root@falcon grub]# ls -la /etc/pki/rpm-gpg
total 80
Why the bloody hell can’t it check that at the start before chewing up 1.5Gb of my monthly bandwidth.
Copied the *KEY* files from the FC10 kickstart directory I had… and rerun the command again.
Will it find the files in cache or will it download them all again, all the rpm files are in cache now.
No, it wanted to download them all again, ctrl-C. Checked the man page again, -C flag seems to be want I want to use cache, try again. Hmm, imported a key this time, must be one of the ones I copied in as there were none in cache. Ok, it seems to be using the cached files.
…
installing package busybox-anaconda-1.10.3-3.fc10 needs 410MB on the /usr filesystem
installing package xorg-x11-fonts-ISO8859-1-100dpi-7.2-6.fc9 needs 410MB on the /usr filesystem
installing package ctags-5.7-3.fc10 needs 410MB on the /usr filesystem
installing package system-config-rootpassword-1.99.4-2.fc10 needs 410MB on the /usr filesystem
installing package wine-core-1.1.7-1.fc10 needs 424MB on the /usr filesystem
installing package jlex-javadoc-1.2.6-6.3.fc10 needs 422MB on the /usr filesystem
installing package xorg-x11-drv-apm-1.2.0-1.fc9 needs 422MB on the /usr filesystem
installing package xorg-x11-xkb-utils-7.2-7.fc10 needs 422MB on the /usr filesystem
installing package sound-theme-freedesktop-0.2-2.fc10 needs 423MB on the /usr filesystem
installing package xorg-x11-drv-neomagic-1.2.0-1.fc9 needs 423MB on the /usr filesystem
…etc…
No error message at the end of the install to say the install failed, but $? was 1… So I must assume it is saying there is no space ???.
[root@falcon usr]# df -k /usr
Filesystem 1K-blocks Used Available Use% Mounted on
/dev/hda6 4956284 4003744 696708 86% /usr
I guess I also assume the –obsoletes option is a waste of time as it’s obviously not taking into acount packages that would be removed.
rpm reports zero fc10 packages installed.
Boot time again, time to run gpartd. /home and /usr partitions are side by side so I’ll move a couple of Gb from /home to /usr.
Ok, thats done, lets see what happens this time then.
[root@falcon ~]# df -k /usr
Filesystem 1K-blocks Used Available Use% Mounted on
/dev/hda6 6912316 4004992 2693340 60% /usr
So here we try again. This time it only found 1.4G of files needed. Still using the same cached files (at least I assume 40% io-wait time indicates that).
Barf. Segmentation fault.
Try again… well yum using 0% cpu and no disk IO is a bad sign, left it 20mins, nothing changed. Reboot -Fr time.
Server back up, try again.
yum –obsoletes –disablerepo *update* upgrade -C
What !!!!
Running Transaction Test
Usage: new-kernel-pkg [-v] [–mkinitrd] [–rminitrd]
[–initrdfile=] [–depmod] [–rmmoddep]
[–kernel-args=] [–remove-args=]
[–banner=] [–multiboot=multiboot]
[–mbargs=mbargs] [–make-default]
(ex: new-kernel-pkg –mkinitrd –depmod –install 2.4.7-2)
error: %unknownscript(kernel-2.6.27.5-117.fc10.i686) scriptlet failed, exit status 1
Finished Transaction Test
Transaction Test Succeeded
Lots of crap in the log like
[(‘installing package k3b-1.0.5-6.fc10 needs 5MB on the /usr filesystem’, (9, ‘/usr’, 4444160L)), (‘installing package planner-0.14.3-6.fc10 needs 7MB on the /usr filesystem’, (9, ‘/usr’, 6684672L)), (‘installing package redhat-lsb-3.2-2.fc10 needs 7MB on the /usr filesystem’, (9, ‘/usr’, 7270400L)), (‘installing package devhelp-0.21-3.fc10 needs 8MB on the /usr filesystem’, (9, ‘/usr’, 8003584L)), (‘installing package kdepim-4.1.2-5.fc10 needs 28MB on the /usr filesystem’, (9, ‘/usr’, 28553216L)),
dah, dah, day. Transaction set succeeded my ass
[root@falcon ~]# rpm -qa | grep -i fc10
Nothing, no matches. Looks like the installation script broke so badly an error message wasn’t even produced.
Yum upgrade -C kernel –disablerepo=*update* again, fails on a hal dependency
yum upgrade -C hal –disablerepo=*update*, fails on a shadow-utils dependency
yum upgrade -C shadow-utils –disablerepo=*update* fails on xenner not available in cache, man page says -C still lets it use the net to resolve dependencies, obviously manpage is wrong
Isn’t yum supposed to resolve dependencies ?.
yum upgrade shadow-utils –disablerepo=*update* from the net repos instead of cache… another package conflict.
About ready to just use rpm to add all the FC10 packages from the kickstart directory with dependency checking disabled, thats the last resort (well ok, last resort is booting off the mkCDrec backups and rolling back to a nice stable FC6 system).
But I’m giving up on attempt 5 for now. Not really sure how to use yum (or rpm) to resolve all the dependency errors that would undoubtably occur from adding all the rpms with dependancy checking turned off, I probably would prefer to roll back to FC6.
Attempt 6 being tried.
Same error as a standalone CD install, but I’ve figured it out now, so the CD install media upgrade may have worked.
Reconfigured my web server as a kickstart server for FC10 (and unplugged it from the internet as I needed to shutdown the firewalls around it do use it as a kickstart server), booted off the install CD with askmethod, typed in the URL. All looking good.
Crap, the same error as doing a standalone CD install, error making directory /mnt/sysimage/var/cache/yum/anaconda-InstallationRepo-200811191949.i386. Now if only I knew where /mnt was.
Booted off CD again and selected boot from local hard drive — failed !. Removed CD and booted from hard drive, worked just fine.
Will create a /sysimage/var/cache/yum (already have a /var/cache/yum… ahh, wait a minute my /var/cache/yum is a symbolic link to a LVM drive with more space, wonder if that is the issue with the install media ?. Probably, bugger why does it need to use cache, my /var is less than 1Gb in size. Moved link aside and created a directory yum, reboot again and see what happens. Be a pain if that is the issue with upgrading from the installmedia as I can’t use gpartd to increase the size of /var, I have LVM partitions in disk slices around it, but FC10 be default in a fresh install by default slaps everything into an LVM filesystem (apart from /boot) so it must know about LVM ?, it obviously doesn’t bother to detect anything but it’s expected defaults, if that.
Hmm, that seems to have been it, got past that error, now it’s at the dependency check stage again. What a useless error message that one was then, /mnt/sysimage/var/cache/yum is /var/cache/yum on the local HD, if it didn’t have /sysimage/ in it I would have figured that out days ago.
Maybe I should have gone for a kickstart install in the first place, but I don’t have servers lying around to put it on. But at least it seems to be installing packages now.
429 of 1257 packages completed, then the finishing upgrade progress bar appeared. Seems like while this has installed something it hasn’t done a full upgrade.
Hmm, moved up to 431 and froze ?. No eventually ‘starting install process’ again. It seems to be handling the lack of space available in /var/cache, I’m actually impressed.
Or maybe not, moved to ‘preparing transaction from installaion source’ and has been sitting there for 20mins. I’m patient, I will give it another two hours. If it hasn’t moved after that it’s probably mkCDrec restore time.
Bugger, it crapped out, dependency issues again, 90% of them on man pages for gods sake.
And on boot up, if had the FC10 kernal, but got selinux denies on the / directory.
And it kept throwing up the full screen x-windows login, not that it worked, just kept dying and restarting.
Booted of the install CD and checked /etc/init.d, default runlevel is 3 so why the full screen login crap ?. Mounting the system in recovery mode, doing the chroot and df -k shows all my LVMs are there.
Ok, about ready to recovery from the backup set now. cd’d to the FC10 package directory and doing a ‘rpm –nodeps -U *’. If it boots up after that I might think about trying to get everything into sync.
If it doesn’t I have no choice but to roll back to FC6.
Ok, it didn’t work. I have a fried server. selinux denies on /, what sort of upgrade is that.
Crap.
Well I must now burn all my mkCDrec CD iso images onto real CD’s and recover back to FC6.
After burning 5 CDs, wait, why don’t I backup and see if a fresh install works first, after trying to remember where all my customised config files are.
Booted off the install CD again into recovery mode, enabling the network. eth0 came up but no network, manually trying to start the network and add routes showed missing modules in directory /var/lib/modules/…fc10.i586. Checked, there was a reason for that, the upgrade had created /var/lib/modules/…fc10.i686; so the upgrade decided to use i686 but put in an i586 kernel ???. Anyway, scp -rp the 686 to the 586 and the network came up, still lots of errors logged but it came up. Changed selinux config to permissive just in case, although I intend to stay in the recover shell until the backups are all done.
Lots of scp -rp’s off the server onto my desktop which fortunately has (had :-).) lots of space.
This is still ongoing, about 100Gb to copy, selectively copy, will take a while.
