Upgrading to Ubuntu Edgy Eft a "Nightmare"

Theovon writes, "It's only been two days since the announcement of the official release of Ubuntu 6.10 (Edgy Eft), and the fallout has been very interesting to watch. By and large, fresh installs of Edgy tend to go well. Many people report improved performance over Dapper, improved stability, better device support, etc. A good showing. But what I find really interesting is the debacle that it has been for people who wanted to do an 'upgrade' from Dapper (6.06). Installing OS upgrades has historically been fraught with problems, but previous Ubuntu releases, many other Linux distros, and MacOS X have done surprisingly well in the recent past. But not Edgy." Read on for the rest of Theovon's detailed report.

Reports are flooding in to Ubuntu's Installation & Upgrades forum from people having myriad problems with their upgrades. One user described it as a 'nightmare.' Users are producing detailed descriptions of problems but getting little help. This thread has mixed reports and is possibly the most interesting read. Many people report that straightforward upgrades of relatively mundane systems go well, but anything the least bit interesting seems not to have been accounted for, like software RAID, custom kernels, and Opera. Even the official upgrade method doesn't work for everyone, including crashes of the upgrade tool in the middle of installing, leaving systems unbootable, no longer recognizing devices (like the console keyboard!), reduced performance, X server crashes, wireless networking problems, the user password no longer working, numerous broken applications, and many even stranger things. Some of this is fairly subjective, with Kubuntu being a bit more problematic than Ubuntu, with reports that Xubuntu seems to have the worst problems, and remote upgrades are something you don't even want to try. Failed upgrades invariably require a complete reinstall. The conclusion from the street, about upgrading to Edgy, is a warning: If you're going to try to take the plunge, be sure to make a backup image of your boot partition before starting the upgrade. Your chances of having the upgrade be a total failure are high. If you're really dead-set on upgrading, you'll save yourself a lot of time and headache by backing up all of your personal files manually and doing a fresh install (don't forget to save your bookmarks!).

Comments

It's been out, what, three days?

[Plug]

Users are producing detailed descriptions of problems but getting little help

I remember rushing to try XGL and Compiz the day they were released, and getting nowhere. About a week later the smart people who do such things had figured it out, and I was able to run it, but it was still pretty 'hardcore' and prone to breakage. About three weeks later it was simple.

Don't upgrade on the first day and expect things to go smoothly. You can only be as good as your last RC, and not enough people upgrade them to be able to find all the bugs. Wait a week and then answers will have been found for all the common problems.

Open source is crying out for more QA people. All you have to do is report a bug, or help by triaging the bugs that are there. It's a contribution that almost anyone can make.

Re:use gentoo and never do another dist upgrade ag

[Anonymous Coward]

use gentoo and never do another dist upgrade again

By the time gentoo is done compiling ubuntu will have released another version with all the bugs fixed.

Don't worry!

[Rik Sweeney]

The problems will all be fixed on Patch Tuesday.

Worked for me and why it happened...

[drgonzo59]

I have been using the development Eft tree ever since they opened it (I like to live on the 'Edge' I guess). I watched new upgrades trickle in over time. The biggest problems were the volumeid changes i.e. referring to the drives using and UUIDs instead of /dev/hd[a-x][0-9] format coupled with a change in udev (and or kernel) that re-mapped the drive order and names. That caused a bit of a headache but I thought it eventually got fixed. Otherwise, there have been no major problems.

The reason I think the upgrade disasters happened is because most developers have been upgrading gradually, over time, just like me. After the release, they assumed upgrading works fine and focused most of the testing on fresh installs. This left the situation of a sudden dist-upgrade from Dapper to Eft un-tested.

In general testing upgrades is pretty difficult. One has to account for X possible previous versions (Dapper, Hoary, Breezy along with mixed software from universe repositories installed by hand) times Y possible hardware configurations. This results in a lot of testing scenarios....

My other take on the situation is that a lot more people are upgrading and therefore there is a total increase in upgrade problems. A year or more ago, there weren't that many Breezy users who upgraded to Dapper (just because there weren't that many Ubuntu users). Now there are a lot more users --- a lot more upgrades --- a lot more upgrade problems.

The change no-one mentioned: bash-dash

[kestasjk]

As a way to get some scripts to execute faster they changed from using bash as the default shell, to dash. dash breaks compatibility all over the place, none of the extensions found in practically every other bourne shell derivative are there. I first found out about this when someone using one of my scripts reported that 'read -s' (for reading passwords without echoing them) and 'trap function SIGINT' both give errors.

So if the scripts you write are going to be used on Eft, you have to either drop a lot of functionality, or tell users to replace #!/bin/sh with #!/bin/bash (which, of course, only works on Eft; it's /usr/bin/bash elsewhere, /usr/local/bin/bash in other places, bash doesn't come on OS X and BSD but /bin/sh works, etc).

A bit of a reckless move for a bit of extra speed. It would have been more respectable if the Ubuntu team had worked on optimizing bash instead of going for a crippled, but faster, shell.

Re:The change no-one mentioned: bash-dash

[ChrisJones]

This is a symptom of a long-standing misunderstanding about shell scripting.
If you have #!/bin/sh you should be using POSIX shell, which will execute fine in bash, dash or the old sh. People run into problems because they've put #!/bin/sh and then used bash-only syntax - ie they should already have used #!/bin/bash, but didn't because they didn't read any docs and don't know better.

Re:The change no-one mentioned: bash-dash

[cortana]

It would be nice if bash rejected the use of (or at least spewed warnings about) bash-extensions when it is invoked as sh.

Re:Yep, bull.

[kjart]

Ubuntu is apt-based. Contrary the the "OS upgrades are typically fraught with trouble" claims of the article, upgrades for debian-like systems are usually flawless -- people do them on a DAILY basis with debian sid and (k)ubuntu's development versions, never mind once every 6 months or so. This article is FUD.

Maybe read the rest of the sentence you quoted: "but previous Ubuntu releases....have done surprisingly well". RTFA is one thing, but Read The Fucking Sentence? Come on.

Also, disagreeing with an article doesn't make it FUD. Perhaps you should tell all the people on the linked to Ubuntu forum that all their upgrades went flawlessly?

Re:Gentoo is why I switched to Ubuntu!

[also-rr]

Gentoo was an even bigger nightmare of manual updating of configuration scripts and bizarre breakages whenever I would do updates. Don't even get me started.

Oh, indeed. I have a

Powerbook, 100% up to date against Edgy Eft. Total time spend fixing upgrade bugs: 5 minutes.
Workstation, 100% up to date against Dapper Drake. Total time spent fixing upgrade bugs: 2 minutes.
Home server, 100% up to date against Gentoo. Total time spent fixing upgrade bugs: 966,352 subjective years.

Despite that there are many reasons to use Gentoo instead of Kubuntu - after all if you wanted the easy life you wouldn't be using Linux in the first place.

dist-upgrade problems

[JRiddell]

Hi, I make Kubuntu. I'm well aware that dist-upgrade has a lot of problems with upgrading to edgy. That's why porting the upgrade tool from Ubuntu will be a priority for Feisty. In the mean time you can use the Ubuntu upgrade tool on Kubuntu fine or you can dist-upgrade and then explicity tell it to install/upgrade the packages it keeps back.

Rethink

[FishandChips]

I'd never consider upgrading a distro like this. Save off your settings and personal files, wipe and reinstall. As many have found, the alternative is asking for trouble.

Even so, let's hope some good comes of this. Perhaps it will encourage the Ubuntu team to take a hard look at what they're doing and where they're at. In retrospect, calling anything like this "Edgy" was a mistake. Ubuntu is aimed at newer and less technically-minded users on the desktop, primarily. That puts a premium on easy, simple and reliable, not on "edgy" as in "the latest gizmos for techies". Techies are not Ubuntu's natural territory. If you want the bleeding edge and all that goes with it, there are 1001 other distros to use. Maybe Ubuntu will decide that its core appeal does not lie in this game, and adjust accordingly. Otherwise, imho, it risks losing the tremendous goodwill it has built up. Ubuntu has never been "just another distro", but if it allows itself to be led only by what developers want, it could easily become one.

Re:interesting

[vhogemann]

Well,

There is this little file called /etc/iftab that handles this kind of issue:

victor@vertigo:~$ cat /etc/iftab
# This file assigns persistent names to network interfaces.
# See iftab(5) for syntax.
 
ath0 mac XX:XX:XX:XX:XX:XX arp 1
eth0 mac XX:XX:XX:XX:XX:XX arp 1

;-)

Re:interesting

[xenocide2]

Submitting a good bug report is a fair amount of work: you need to check for dupes, lay out the conditions necessary to reproduce it, give details like your hardware, etc. This is a lot more work than the work necessary to, say Google something, or read a man page, yet we already know that this is too much to expect from most users. If Ubuntu wants things to be easy enough for people that don't know how to google something, then they cannot reasonably expect everyone to submit bug reports.

This is a perception that we need to try very hard to dispel. The most important aspect of a bug tracker is bring people together in one place. A bug is a jigsaw puzzle, with different people having different parts of the puzzle. Some people find ways to trigger the bug, others find ways to accomplish the same thing without triggering the bug, some people fix the flaw in the source code that caused the bug, and some people find how other people fixed the bug. We absolutely need as many people as possible coming together to solve a given bug, in hopes of finding the right combination of the above sets. What we find is that people are perfectly willing to write nasty things on a forum, but for some reason they won't put in the same level of effort into a bug report. This effectively divides that community we needed to build, where people who find bugs complain in one place, people who come up with workarounds and find patches in another, and programmers hiding elsewhere. Writing software to find probable duplicate bug reports should not be a significant challenge in 2006. You said yourself, Google is a good tool used to find how other people solved a bug. It stands to reason that much of the same technology can be applied within say, Launchpad. If writing good bug reports is too hard, we should find ways to make it easier, or find ways to use "bad" bug reports, rather than let everyone give up in isolated desperation.