Would you to the dist-upgrade command in the first week this is released?

Submitted by Nelson_Y on Sun 15 Feb 2009

Tags: , ,

 

Yes, I am confident that this will work perfectly!  <-> 38%832 votes
Yes, I will do - but I do not advice you if you are not expert Admininstrator !  <-> 12%276 votes
Yes, but only if the software I will use were tested by myself, and I am sure it works !  <-> 6%144 votes
No, better wait a couple of days.  <-> 8%177 votes
No, better wait for some weeks.  <-> 16%367 votes
No, better reinstall the system from 'zero'.  <-> 7%168 votes
No, better wait some months.  <-> 4%101 votes
No, 4.0r6 is stable and will remain 'stable' for at least a year yet.  <-> 5%115 votes
Total 2183 votes

Posted by ajt (195.112.xx.xx) on Sun 15 Feb 2009 at 17:52
[ Send Message | View Weblogs ]

I upgraded and everything when in okay. I'm happy.

No, better reinstall the system from 'zero'.

What goes wrong with operating systems that mean that they need periodic reinstalls and upgrades are impossible? At work last week the Windows admins were talking about reinstalling a server because it was broken, what's scary is that pretty much nothing has happened to it in years - it's not as if it has a lot of users or lots of applications being added or removed. How can it go so badly wrong just by running? I know it's not just a Windows problems, lots of Linux users seem to reinstall all the time - why?

--
"It's Not Magic, It's Work"
Adam

[ Parent ]

Posted by Anonymous (82.29.xx.xx) on Mon 16 Feb 2009 at 01:09
Windows often needs reinstalling when a problem develops. I had an update go wrong, the registry corrupted, and there was no way to backout to a consistent set of registry values and installed software. Also seen it chew up configuration database for IIS too often. Older versions of MS SQL just occasionally go crazy and eat their data. I think ultimately too much data in too few files, systems that are closed both to user and admin so the admin doesn't know what happened let alone how to fix it, poor software quality, and poor design decisions. Some Linux distros do some of the above, but mostly I think it is ignorance, and experience of Windows where reinstalls can often solve problems for mysterious reasons.

Debian seems almost indestructible if you stay with the official packages.

[ Parent ]

Posted by ajt (204.193.xx.xx) on Mon 16 Feb 2009 at 12:50
[ Send Message | View Weblogs ]

The Windows Registry does seem to be a fragile beast, with little reputable documentation on how to fix it safely in the public domain. There are lots of myths and cults surrounding the Registry though...!

I sometimes find that the Windows approach of "reinstall to fix" mentality does seem to be leaking into Linux. Distros popular with new users fleeing from Windows appear to bring this "reinstall to fix" approach with them. A cynic would say it's because it's easier to reinstall than it is to understand the problem and fix it, giving the benefit of the doubt you could say it's not possible to fix on Windows so the user doesn't realise it is probably fixable on Linux.

I must confess that other than messing with my experimental first installs of Debian, I've never needed to reinstall Debian to fix a problem, nor do I reinstall package - unless I've manually screwed up package...

Stable does indeed seem to be bullet-proof and the upgrades do just work.

--
"It's Not Magic, It's Work"
Adam

[ Parent ]

Posted by mcortese (213.70.xx.xx) on Thu 19 Feb 2009 at 12:25
[ Send Message | View Weblogs ]
The Windows Registry does seem to be a fragile beast
Yes, and I wonder if the gconf system isn't following the same path. That's so far from the UNIX style!

[ Parent ]

Posted by ajt (204.193.xx.xx) on Thu 19 Feb 2009 at 14:01
[ Send Message | View Weblogs ]

A binary configuration database or registry isn't such a bad idea, if it's very fast, very robust and easy to use, as long as there is some way of easily maintaining and repairing it. So you can say something like etc-recompile and it rebuilds a binary database for you of all your normal text files in the /etc directory. However it would require that all the applications to be changed to use the new configuration engine - so I doubt it will ever happen.

The problem with the Microsoft registry is that it's unreliable with no easy way to maintain. Also overtime it also gets bloated and slow as it becomes "fragmented". It's probably not a good example of the concept.

--
"It's Not Magic, It's Work"
Adam

[ Parent ]

Posted by mcortese (213.70.xx.xx) on Fri 20 Feb 2009 at 16:42
[ Send Message | View Weblogs ]
Agreed.

My worries are about gconf, which is not binary (thus does not gain in speed) and not so easy to maintain, so it reaches none of the objectives that you outlined in your post. Instead I often find it clogged with unused keys inherited from older version of some software, sometimes remnants of the gconf-to-gconf2 transition.

[ Parent ]

Posted by Anonymous (83.33.xx.xx) on Thu 19 Feb 2009 at 22:21
i agree on this.

[ Parent ]

Posted by mario (201.210.xx.xx) on Sun 15 Feb 2009 at 20:30
[ Send Message | View Weblogs ]
I'll first update my VMs and then the server.

[ Parent ]

Posted by Anonymous (75.182.xx.xx) on Fri 20 Feb 2009 at 15:37
If you are using Xen and upgrading from Etch to Lenny, learn from my mistake.

I dist-upgraded the VMs without much trouble. However, when I did `aptitude upgrade` on the Xen hypervisor it was still holding several packages back. I then foolishly did a dist-upgrade anyway.

I then had a system with a version of Xen that depended heavily on Python 2.4, but also had now upgraded to Python 2.5. In addition, to find out about that problem, I had to roll back to the previous kernel, the newest one with Xen wouldn't even boot (on my Dell PowerEdge 600SC mini-tower).

Finally I had to rescue the whole thing using Knoppix and bare metal restore over the network. This was my dev machine, and I basically had to rsync (over ssh) my matching production machine's OS installation over the network and nuke the mess I had made on my dev dom0. I then restored key files like /etc/fstab, /etc/udev/rules.d/, /etc/hostname, /etc/hosts, and so on.

Amazingly, I got it all working again. If you don't have 100% coverage of every single file in your backups or a matching dom0 installation you can use as the basis of a new installation, I would leave those packages held back and not be as aggressive as I was.

I still love Debian and Xen. The fact that you can do a bare metal restore of an entirely different operating system installation securely from a running server 2500 miles away is a testament to Linux. Still, it's nicer to just run the updates and not have such excitement.

[ Parent ]

Posted by Anonymous (74.10.xx.xx) on Fri 6 Mar 2009 at 21:25
If you follow the release notes at http://www.debian.org/releases/lenny/releasenotes it warns about clearing out non-upgradeable packages before dist-upgrade. I was running xen-hypervisor on etch and successfully upgraded to lenny. Took me two tries, because the first time, I did what you did. The second time a friend of mine bashed me over the head with the release notes URL and all was well. For example, the release notes say it is best to aptitude install apt aptitude before doing dist-upgrade, because the new aptitude has better resolution handling that leads to a safer upgrade.

[ Parent ]

Posted by Anonymous (24.36.xx.xx) on Sun 15 Feb 2009 at 21:31
I would, and have complete confidence in doing so, but I have been running a mix of Testing (Lenny) and Unstable for almost a year already. I have had no issues with it. I am going to wait for a few weeks for Squeeze to settle down and will then look to selectively upgrading packages for this system at that point.

Thank you to all of those involved with creating such a great distribution, I am thrilled to be a user.

[ Parent ]

Posted by Utumno (60.248.xx.xx) on Mon 16 Feb 2009 at 02:55
[ Send Message | View Utumno's Scratchpad | View Weblogs ]

On my desktop , I haven't reinstalled since early Sarge. dist-upgrade has always worked for me.

[ Parent ]

Posted by Anonymous (207.178.xx.xx) on Mon 16 Feb 2009 at 16:37
I've been running Lenny on two workstations for a few months, one of them a fresh install (new box) and one of them a box that was installed as Sarge, upgraded to Etch, then upgraded to Lenny three weeks ago.

The upgrade from Etch to Lenny is much simpler and smoother than Sarge to Etch. (That was complicated by the transition from hotplug to udev).

If you're not already running Lenny, I see no reason not to upgrade today. A 22-month development and testing cycle ought to be enough.

[ Parent ]

Posted by Anonymous (134.84.xx.xx) on Mon 16 Feb 2009 at 17:06
I'm tracking testing so the upgrade didn't seem that difficult.

[ Parent ]

Posted by mcortese (213.70.xx.xx) on Thu 19 Feb 2009 at 12:17
[ Send Message | View Weblogs ]

So what you actually did was upgrading from lenny (former testing) to squeeze (current testing). Isn't it?

[ Parent ]

Posted by Anonymous (81.149.xx.xx) on Mon 16 Feb 2009 at 17:45
I did this for the first time ever yesterday. The new kernel wouldn't boot because the scripts in local-top were executed in the wrong order. Quite an achievement as there are only three of them and one of them doesn't do anything on my systems.

Ahem.

[ Parent ]

Posted by Anonymous (81.149.xx.xx) on Wed 18 Feb 2009 at 17:51
The ping and traceroute utilities are broken on AMD64 (Opteron). The times returned are garbage. I compiled every release back to ping 1.4.2 on the Opteron box with the same results, so it's probably something on which these utilities rely, not the utilities themselves.

Both seem to work on Athlon and i386.

[ Parent ]

Posted by Anonymous (207.112.xx.xx) on Tue 17 Feb 2009 at 13:00
I stopped using dist-upgrade when I noticed that a freshly installed Debian/etch was different then an upgraded sarge to etch.

[ Parent ]

Posted by Anonymous (213.227.xx.xx) on Wed 18 Feb 2009 at 11:55
It depends how you upgraded. �re you talking about obsoletes? What aptitude command did you use? How is it different?

[ Parent ]

Posted by lykwydchykyn (72.237.xx.xx) on Tue 17 Feb 2009 at 16:31
[ Send Message | View Weblogs ]
I've started upgrading the etch boxes at work; so far this upgrade looks a bit uglier than past upgrades. I mean, aptitude is throwing more of a fit with complex resolutions. I may hold off on the mission critical servers for a bit, I'll have to make backup images and do some testing.

[ Parent ]

Posted by Anonymous (193.203.xx.xx) on Wed 25 Feb 2009 at 11:55
Strictly speaking this should be tested anyway.. My experience though is the opposite.. Upgrades to lenny from etch are a damned sight easier than those from sarge to etch on complicated setups...

[ Parent ]

Posted by paquin (87.14.xx.xx) on Wed 18 Feb 2009 at 03:10
[ Send Message | View Weblogs ]
Great work! Upgraded all my servers and workstations from etch to lenny (i386 and AMD64). Easy as drink a cup of tea. Just for fun also my personal Netbook (Acer One) is currently running with current stable release...

paquin

[ Parent ]

Posted by Anonymous (207.134.xx.xx) on Wed 18 Feb 2009 at 13:07
Is it "would you TO" or "would you DO"?

[ Parent ]

Posted by lionslair (203.161.xx.xx) on Thu 19 Feb 2009 at 09:28
[ Send Message ]
I upgrade using aplitude and it went all smooth. I got some small problem that I can't access my samba shares but I will get it sorted. Apart from that nothing broke.

[ Parent ]

Posted by Anonymous (94.189.xx.xx) on Thu 19 Feb 2009 at 23:28
I would wait some time. I reported a bug concerning Xen on Lenny and got ZERO feedback on that. I will test it on new "stable" Lenny in a few days and see what happens. But I really wonder, how did those 100+ bugs got fixed in one (last) week before the release?

[ Parent ]

Posted by simonw (84.45.xx.xx) on Wed 25 Feb 2009 at 08:29
[ Send Message | View Weblogs ]
Voodoo...

Depends on the bug - but someone does look at each of them and make the decision "live with it", or "fix it", hence the "<Release Name>-ignore" tags.

One has to appreciate that all general purpose operating systems (except perhaps some safety critical ones) release with a lot of "release critical bugs" or equivalent issues.

Lenny has had a substantially lower release critical bug count than Etch since around January 2008, at some point one has to make the call that Lenny is simply better.

Release critical can include non-technical issues, such as failure to comply with DFSG. DFSG issues often occur in all released versions of the same software, Linux being the controversial case. Or issues that affect a package on one architecture.

[ Parent ]

Posted by e5z8652 (206.174.xx.xx) on Fri 20 Feb 2009 at 04:31
[ Send Message | View Weblogs ]
I did wait for 24 hours, but since Monday was a holiday and there was very little load on my servers I just did it.

Almost everything came up roses. The only problem I had was with ucarp. Had to tweak some things to get that working again.

[ Parent ]

Posted by mario (201.210.xx.xx) on Sat 21 Feb 2009 at 06:26
[ Send Message | View Weblogs ]
I had a problem with a script that used php4-mysql, it seems php5 removed it to install php5-mysql, and openntpd stopped working, I had to remove and install it again (ntpd) and it worked ok.

[ Parent ]

Posted by Anonymous (80.85.xx.xx) on Fri 20 Feb 2009 at 16:18
I upgraded.
I used before apache 1.3 and php4. It was problem, and some work to use apache2 and php5.
When I upgraded, the installer erased user "spamd" from /etc/passwd, and after the exim4 (with spamassassin) did not start. I had add user "spamd" again manualy.

Anything was nice.

By

Totyi

[ Parent ]

Posted by Anonymous (71.94.xx.xx) on Sat 21 Feb 2009 at 16:42
Are you asking would I do a dist-upgrade to Lenny? Sid? What does "this" mean?

[ Parent ]

Posted by Anonymous (193.136.xx.xx) on Fri 27 Feb 2009 at 00:10
After getting the Exim4 configuration all messed up and the /etc/sysctl.conf settings ignored from the simple 'apt-get upgrade' to Lenny, I'm not using dist-upgrade any time soon.

[ Parent ]

Posted by Anonymous (80.101.xx.xx) on Fri 27 Feb 2009 at 12:39
I'm upgrading. Most of our systems are LAMP boxes, using dotdeb.org packages. (http://www.dotdeb.org/) As a consequence I didn't have to upgrade PHP5 or MySQL. Therefore I knew the applications running on top of PHP would still work.

The only problem I had was my bnx2 network card not working on a Xen host (see http://techblog.procurios.nl/k/news/view/16583/14863/ for my story)

Still, I started the least critical server. Testing our applications first, then working my way up to essential mission critical servers.

[ Parent ]

Posted by Anonymous (88.86.xx.xx) on Sun 1 Mar 2009 at 08:19
NIX never screws up (oh if you don't use some crap distrib like suse or fedora which m8 have to be reinstalled just as often as a windows box has to be done but you all know that). Debian is an excellent distribution. For newcomers the only thing what you have to watch is the syntax changes in the config files.

I had many servers which are still running with the exact same kernel+exact same packages what I configured them with at the first time. For this hobbylinux community here it m8 be hard to understand what production version of OS/Software means. If you don't have _reason_ to change something you won't.

And about my desktop installs, I had the same debian install for 4 years what I just upgraded constantly, then have the same ubuntu install since 3 years.

[ Parent ]

Posted by Anonymous (153.107.xx.xx) on Mon 2 Mar 2009 at 23:25
The 4 "real" servers I maintain are still on Etch. I would have done all 4 by now normally, but they are all in places I can't really do anything during the day... And I don't have physical access during the night.

I don't like remote upgrades when I don't at least have a key to the building, or someone who can get in there and be my eyes.

[ Parent ]

Posted by Anonymous (81.136.xx.xx) on Fri 6 Mar 2009 at 21:11
Yeah, I hit "no, 4.0r6..." too. But it's because I have servers out there that _must_ _not_ go down, tbh, even rebooting them would be a problem, I don't do anything needing a reboot unless there's a security patch that effects us.

Think I'd be pretty confident on workstations and possibly the office servers though

[ Parent ]

Posted by Anonymous (82.10.xx.xx) on Thu 5 Mar 2009 at 21:59
No, I nearly went ahead and as there is no gallery2 package in Lenny (Wassat all about?) it tried to uninstall gallery2. Duh. Come on Debian?!

Peter

[ Parent ]

Posted by Anonymous (84.238.xx.xx) on Fri 6 Mar 2009 at 21:17
My first Debian was woody. Since then the system gets just dist-upgraded. It works like a charm with minor cases where i got broken packages and i fixed the things manually.

[ Parent ]

Posted by Anonymous (67.182.xx.xx) on Thu 12 Mar 2009 at 18:56
Lots of people seem to report no problems, but lots of people also report having problems. Unfortunately, I used to run Fedora (and RedHat before that) as my distro of choice before I started using Debian. I am somewhat jaded as every upgrade in Fedora gave me issues, and it was actually easier to just do a reinstall (less time, headaches). However, I did learn a lot from fixing problems when things went wrong. Maybe that's a good thing??? Basically, I just use LVM and keep my /home partition (I also don't use /usr/local, but /home/apps just as a quirk to get around any problems.) So, usually I can get a fresh install working in just a couple of hours. However, I am feeling more confident that I might be able to survive an upgrade with Debian. Maybe for the next release after Squeeze, I'll give it a try. However, I will always prefer to spend an hour installing fresh packages instead of fixing problems with upgraded ones... For servers, I typically never upgrade, and just start building everything myself from source. I'm still running my Fedora 5 box, and it's doing just fine for me. However, I would like that one to be Debian as well... hmmm... Ok, I'll stop rambling now! :)

[ Parent ]