Fedora 20 to Fedora 21 upgrade woes
This upgrade did not go nearly as smoothly as the last one, mostly because NVIDIA’s 343.22 driver no longer supports the 8200 integrated video on my Asus motherboard. After several failed attempts to revert to the nouveau driver or the 340 NVIDIA driver, I decided to find a newer video card. I picked up a PNY 720 card at a Fry’s while on business travel. Everything pretty well works except that I clobbered initramfs-fedup.img in the process. This was resolved when the kernel was updated to 3.17.7-300.fc21. Samba was a bit tricky to get working again. I tweaked smb.conf slightly, and ran these commands:
setsebool -P samba_domain_controller on
setsebool -P samba_enable_home_dirs on
uncommented ‘netbios name’, removed the IP range restriction in the ‘inferfaces line, and uncommented these lines in smb.conf:
local master = no
os level = 33
preferred master = yes
then restarted via:
systemctl restart smb.service
systemctl restart nmb.service
Which all seemed to work once, then failed on the next reboot.
Also, I still have trouble booting. Fedup somehow created a second /boot as a directory on Logical Volume 0, (/), rather than mounting the boot partition via the specified UUID. I’m running hardware RAID-1 mirroring on an ASUS M3N78-VM with an NVIDIA chipset. Something is confusing the kernel during boot time. Some of the key journalctl -xb messages are (timestamps and hostname removed):
systemd-sysv-generator[757]: Could not find init script for httpd.service
(repeat for every possible service)
kernel: ACPI Warning: SystemIO range 0x0000000000000700-0x000000000000073f conflicts with OpRegion 0x0000000000000700-0x000000000000073f (\SM00) (20140724/utaddress-258)
kernel: EXT4-fs (dm-1): mounted filesystem with ordered data mode. Opts: (null)
kernel: nvidia: module license ‘NVIDIA’ taints kernel.
kernel: Disabling lock debugging due to kernel taint
kernel: nvidia: module verification failed: signature and/or required key missing – tainting kernel
systemd-udevd[824]: starting version 217
systemd-udevd[847]: Error calling EVIOCSKEYCODE on device node ‘/dev/input/event2’ (scan code 0xc0183, key code 226): Invalid argument
systemd-udevd[847]: Error calling EVIOCSKEYCODE on device node ‘/dev/input/event2’ (scan code 0xc1005, key code 212): Invalid argument
Etc.
systemd-fsck[927]: /dev/mapper/nvidia_bbhdeagj1: clean, 359/128016 files, 157368/512000 blocks
systemd[1]: Job dev-mapper-vg_zzzzz\x2dLogVol03.device/start timed out.
systemd[1]: Timed out waiting for device dev-mapper-vg_zzzzz\x2dLogVol03.device.
systemd[1]: Dependency failed for File System Check on /dev/mapper/vg_zzzzz-LogVol03.
— Subject: Unit systemd-fsck@dev-mapper-vg_zzzzz\x2dLogVol03.service has failed
systemd[1]: Dependency failed for File System Check on /dev/mapper/vg_zzzzz-LogVol02.
— Subject: Unit systemd-fsck@dev-mapper-vg_zzzzz\x2dLogVol02.service has failed
— Defined-By: systemd
— Support: http://lists.freedesktop.org/mailman/listinfo/systemd-devel
— Unit systemd-fsck@dev-mapper-vg_zzzzz\x2dLogVol02.service has failed.
systemd[1]: Job dev-mapper-vg_zzzzz\x2dLogVol01.device/start timed out.
systemd[1]: Timed out waiting for device dev-mapper-vg_zzzzz\x2dLogVol01.device.
— Defined-By: systemd
— Support: http://lists.freedesktop.org/mailman/listinfo/systemd-devel
systemd[1]: Dependency failed for /home.
— Defined-By: systemd
— Support: http://lists.freedesktop.org/mailman/listinfo/systemd-devel
systemd[1]: Dependency failed for File System Check on /dev/mapper/vg_zzzzzz-LogVol01.
— Subject: Unit systemd-fsck@dev-mapper-vg_zzzzz\x2dLogVol01.service has failed
— Defined-By: systemd
— Support: http://lists.freedesktop.org/mailman/listinfo/systemd-devel
— Unit systemd-fsck@dev-mapper-vg_zzzzz\x2dLogVol01.service has failed.
systemd[1]: Job dev-mapper-nvidia_bbhdeagjp2.device/start timed out.
systemd[1]: rpc-statd.service: control process exited, code=exited status=1
logging in at the recovery prompt and running the following commands:
lvchange -aay vg_zzzzz/LogVol01
lvchange -aay vg_zzzzz/LogVol02
lvchange -aay vg_zzzzz/LogVol03
I also changed fstab to mount /dev/mapper/nvidia_bbhdeagj1 as /boot instead of using the UUID.
^D then boots up properly.
Which only worked once. 🙁 Next reboot would not mount the logical volumes by any method.
I’ve tried running dracut to rebuild the intramfs files, to no avail. I tried re-running grub-install because there was an issue with it, and now it won’t boot at all.
I found that the only way to boot from my SATA DVD is to unplug the RAID drives, which is where the OS lives so that’s pretty useless.
Booting from a usb stick written with liveusb per the Fedora instructions only works if you know to hit the ‘TAB’ key at the “…COM32R…” boot: prompt, then type linux0 .
After re-re-installing grub2, and a few reboots with many odd errors, and setting “firstboot”, followed by 2 reboots and more futzing with lvchange to get drives mounted, it finally booted. Question is, will it boot the next time? Answer: No, had to revert to the FC20 kernel again. 🙁
23-March-2015: Now I seem to have to run dmraid -ay to activate the LVM partitions, followed by the lvchange commands to mount them.
