Paolo Galtieri writes:
I tried to upgrade one of my systems to F26 according to the
documentation at
https://fedoraproject.org/wiki/DNF_system_upgrade
I did the download and everything went fine. I then did
sudo dnf system-upgrade reboot
and here is where things went weird, there was no output at all on the
monitor. The disk light was flashing, so something was going on, but
nothing showed up on the monitor. After a while the disk light stopped
flashing, and I assumed it was time to reboot. I rebooted, but nothing
A successful upgrade reboots automatically. You never need to manually force-
reboot an upgrade.
Anyone know what I need to do to fix this system? There are 2 other
systems
It is certainly possible that you hit some kind of a bug that prevented the
upgrade from either starting or successfully completing.
But by force-rebooting you just made things worse. All signs are that the
upgrade was in progress, but by force-rebooting you ended up with a
completely messed-up, half-upgraded system. Experiences vary wildly, but,
for example, one of my servers does not use rhgb, so the normal boot
environment is text-only, and system upgrades are the same. However with
that server's hardware text output to the system console is slow, and the
last part of a system upgrade runs for about an hour with nothing happening
but "Verifying <package name>" slowly being printed to the console, one by
one for every one of the two thousand-odd packages that were just upgraded.
No disk activity.
Of course, the situation you described is slightly different - no terminal
output whatsoever. But the point is that there are times during a large
system upgrade that pass by with no apparent disk activity. I would not
attempt to interrupt an upgrade for at least 3-4 hours, after seeing nothing
happening.
Anyway, if something truly prevented the upgrade from getting started, I
would've expected little or no changes to have actually occured, and the
system being mostly undamaged; but based on your description, it's very
likely that you interrupted a full system upgrade in the middle, with most
packages being half installed, or uninstalled. Basically, the whole system
is hosed. Unfortunately there is no magic button anywhere someone can push,
and fix a busted upgrade of this kind. Botched upgrades can often be
recovered, by carefully investigating the state of the system, by using an
install image to boot into recovery, and then picking apart the flaming
wreckage. Once I accidentally SIGKILLed a regular "dnf upgrade". Not a full
system upgrade, but a fairly large update, with several hundred packages. It
was a mess, but it was recoverable.
Unfortunately, as I said, there's no recipe for recovering from this kind of
a botched system. The recovery process involves direct examination of the
state of the crapped out system, and figuring out how to fix it.
I want to upgrade, but after this experience I'm hesitant to go
ahead with
the others. This system is my test system so there's nothing important on
it.
Any help is appreciated.
I would suggest wiping the test system, installing F25, installing all
updates, then once against attempting an system upgrade to F26, but keeping
careful notes this time.
I do concur that F26 turned out to be one of those occasional, painful
upgrades. The new version of qemu has some kind of a bug that fraks up guest
VM reboots, causing my Windows 10 guests to reboot into a bizarre recovery
mode, then claiming that the virtual disk is completely corrupted. Which is
a bald-faced lie, there is nothing wrong with it. But I didn't figure it out
until I already wiped one of my Windows 10 guest images, and reinstalled it
from scratch. Still rather teed off, on that account. But that's what's life
on the bleeding edge is all about…