Gentoo not as reliable as before...

[Lokheed]

[Gherald]

[Tamnir]

Hello all,

First a quick introduction, before I get discarded as some crazy CFLAGS-maniac overclocker : a friend and I are running an IT company in Japan, and we are using Gentoo in production environments, including for customers. Our main business is custom software development, and we also actively participate in the open source community, by contributing code and testing time to various projects (specially regarding Japanese support), and more generally by being Open Source Software advocates. Sorry for this being a long post, but I hope you will make it to the end

I would like to start by thanking all the Gentoo developers for their work. They have put together a great distribution, and they are making our work easier every day Now I am afraid but I must agree with crohnnie: I too have seen a noticeable degradation in the quality of the portage tree stable branch in the past few months. "emerge world" used to work like a charm. Our machines use an autosync script that builds binary packages at night, and all we have to do is an "emerge -avK ..." to get an almost instant upgrade. As of late, every once in a while, something fails, and we have to spend time figuring the problem out and restart the emerge manually. While for the Gentoo hobbyist that may be an acceptable small annoyance, for us it more of a concern. Please read on to find out exactly why.

(If it was not obvious yet: no we are not using crazy CFLAGS or anything. We are talking production machines here: our settings are very conservative; and yes, we always etc-update; and we revdev-rebuild too.)

We started using Gentoo about a couple of years ago, and quickly fell in love with it. First I was just using it on one of my workstations, and as I saw the power of portage and the ease of adminitrating a Gentoo box, little by little, we ended up migrating all our machines to Gentoo, and also recommending it to our customers. We are currently running Gentoo on 3 laptops, 4 workstations and 5 productions servers. 3 of these machines are customers'.

Now the important thing is: why did we choose to install Gentoo on all these machines? The three main reasons are:

1) Fast updates. We used to run Debian or TurboLinux. The problem with TurboLinux was that they were too slow to release security fixes. Last year when a serious SSH vulnerability was discovered, it took them like a week to release an RPM with the fix. This week of being vulnerable made our customers very nervious. On the other hand, our Gentoo boxes were already fixed even before we heard about the vulnerability.

2) Latest packages. Now with Debian the problem was different: sure it is very stable, but it is too conservative for us. With the applications we develop, we need some packages that were not even in Debian unstable. So we ended up having to install them manually, and therefore maintain them manually, which quickly became a real drag.

3) Easy maintenance. Gentoo alleviated both the problems in 1) and 2), while maintaining the same ease of maintenance as Debian with "sync && emerge -uD world".

So, rock solid, latest packages, fast updates and a breeze to maintain, Gentoo was like a Linux dream. And that's why it became our distribution of choice.

Now let's go back to point 3: it is really a major point for us. We do not get paid for administrating machines but for developing software. We are a small team and cannot afford to waste man-hours. So it is very important for us to keep system administration overhead to a minimum, since every hour we spend there is an hour lost. That is why we use scripts to automate most maintenance tasks. But unfortunately, the number of hours lost to system administration has noticeably increased lately. Some examples of the problems we ran into: giflib and libungif claiming the same library, need to run fix_libtool_files.sh after a gcc upgrade, and various packages suddently blocking each other. While it is still within an acceptable realm, if things continue going that way, we may have to reconsider using and recommending Gentoo.

Now do not get me wrong: I am not saying the situation is unbearable yet. Gentoo is still our favorite distribution, and Gentoo stable, the system itself, is still rock solid. But it is Portage and upgrades that fail more and more often lately, forcing us to increase our system administration overhead. We have noticed this trend, and apparently we are not the only ones. So maybe there is actually a problem worth looking into. We just thought we should bring this matter to the attention of the Gentoo developers.

Sorry again for the long post, and thank you for reading this far.


Best regards,

--
Yves-Eric Martin

[ciaranm]

[Arker]

These packages that are breaking, I'm curious about. Are they all (or at least a majority of them) related to each other? For example, is it GNOME updates that break? KDE? I know GNOME has been a little shaky for the past year maybe due to API and technology changes. Stuff like that should not really be in the stable tree, but if they aren't, then the devs get flak from the folks who say the stable tree is too old.

~djc
_________________
*LIK*

My other computer is your Windows box.

[jshanab]

I have been using gentoo for over a year and when I first started I had arch set to ~x86. No manual etc updates and no crashes. Then suddely, yes 6 monthes ago it started to go nuts, every emerge world crshed system.

I think it was a combination of my adding more packages and the increase in the number of developers, some who took the ~use flag to mean more of an alpha test than a beta test.

I changed to x86 stable, which took a weekend and was not straight forward, but as fixes catch up with the files that couldn't downgrade it has once again become stable as a rock.

Forum and install files are ahead of documentation, I have now discovered to investigate or take the defaults in files over advice to change them in the guides. Once they started the interactive etc-update, it has now become mandatory on files I've altered in order to keep settings for my system, when it was automatic, it seemed to default to keeping my changes, now the default wipes my changes.

This is very annoying. I am not sure why etc-update can't search for which video card I have chosen or which login manager and keep my settings instead of wiping out my video card.

2 recommendations. Make a copy of important files in .etc before running etc-update and reboot periodically. I know, that last one is kinda funny but I found I have a tendency to leave my computer on 24-7. I hardley ever reboot. What happens to me is something in the update process does in some critical setting that won't take effect unless I reboot. Then weeks, dare I say monthes, later when adobe acrobat crashes X and I end up rebooting, I am fixing a problem whos source was weeks ago.

[mcspiff]

ciaranm, depending on what your trying to build, ive seen maybe a half dozen errors. Some of these packages may be a little exotic, but nothing that wasnt in portage. It happens. Usually gets fixed. My cflags were straight from the handbook.

No, after using gentoo for several years, it just hadnt reached a point where i could be productive on it. Nice for a hobby. I tend to install at the start of summer, whipe it all off in the fall when school comes around. To be honest, i have everything from both linux and windows running like a dream in XP Pro.

Perhaps when their is a stable branch, like that of debian, to build off of ill return to gentoo. But ill keep posting in these threads (and using my gentoo account on a friends box).

[Scottaroo]

Greetings:

I would have to agree that the AMD64 seems a little less stable than the x86 branch. Nothing horrible, but annoying. So as not to have people question my CFLAG choices:

CFLAGS="-march=k8 -O2 -fweb -frename-registers -fomit-frame-pointer -pipe"

The box was built 9/4/2004. It's stable AMD64. Here is some of the weirdness:

EVERY time gnome is updated, my file associations get trashed. I have to go through and regenerate each one of them every time.
Clicking on a media link (sound, video) from firefox causes firefox to segfault.
I don't recall the exact number of packages that had trouble building, but kerberos had trouble repeatedly. There were a few that didn't build though. Since AMD64's default compiler is GCC34, there were issues with packages that didn't want to build because of the compiler.
Don't even get me started on the 64-bit ATI X drivers.
It won't boot if I leave my Battlefield:Vietnam disk in the USB CDROM drive after using XP to play games for a while.

Nothing major, just weird and annoying. I've been thinking about rebuilding the box to see if some of the problems get cleaned up, but it's a lot of work and I have no reason to assume that it would help.

Just my .02
_________________
If Microsoft is your answer, you obviously did not understand the question.

[robet]

I think -fweb is bad. Do a revdep-rebuild too.
_________________
no software patents in europe! | adopt an unanswered post.

[Redeeman]

i would say the oposite... i have been using gentoo for quite some time now, and never have things been so smooth, it is in fact now possible to do a install, without ever having to see a compile fail or anything.. this wasnt possible(for me atleast) back in the 1.4 days....

[loki99]

i have been using gentoo as my one and only os for almost a year now. and every single problem that i ran into, originated from mistakes i made. with two exceptions: updating portage-2.0.51, which was fixed within hours, and freetype giving me troubles with firefox, which seems to be related to my ati-card though!

so, i can't really complain. its the first time, that an os stayed on my hd for so long, without a reinstall.

but posts like Tanmir's shouldn't be ignored either! he and his arguments seem to be quite reasonable to me.

[syg00]

I'll vote for the developers.
Must be a prick trying to deal with all (well most) of the myriad hardware combinations possible.
Not to mention all the crazy bastards using said - present company excluded of course

I think it's fantastic folks like Tamnir push linux commercially - broken systems are hardly likely to endear them to customers.
Maybe better to be broken for a while than compromised for an extended period.
Well informed customers would probably be prepared to pay for the admin time expended - with just a little arm twisiting.

[halz]

I haven't had much success doing emerge world after long periods of idle time.. I left my happy-as-a-clam Gentoo system off for about 6 months, turned it on 2 months ago, emerged world only to witness the world come crashing down.. Fixed all of the little things, and its happy again. A month later, emerge world spews out some more errors on innocent little packages such as xmms, and an Xorg upgrade disconnected my graphical view of the system..

Granted, I have not been able to keep up with how quickly the documention cycles as things change and the number of options increases. I've had success just jumping into the 'install guide' and pretend I am re-installing my system from scratch when I feel out-of-it..

Edit: and now its back to being happy as a clam

[mikecore]

The origanl poster is correct I also noticed that things changed a few months ago.

I use few use flags(gnome, xmms, dvd). and stable kernel. nothing has changed
on my side and since a few months back everytime I update "world" somthing is "Blocking" or somthing fails to compile, and i have to look up the fix. (I also have not installed any new programs)

Im NOT complaining just stating the facts about what I have seen over the last few months! on my system.

[yokem55]

Those that are running 90nm Athlon64's, and are having a lot of compile failures (ending in a segmentation fault) might want to check and see what thier bios is reporting as thier cpu voltage. If it is slightly below what it is supposed to be (ie. 1.38v on a 1.4v chip), try raising the cpu voltage a single notch (1.425 should be the next step up). If your memory isn't perfectly compatible with the A64's on die controller, this slight undervolting might be causing the segfaults. A good test is the mprime program that is in the sci-mathematics/gimps package. If when you run the "torture test" (option 17), and it dies within a few minutes and your cpu is slightly undervolted, you likely have this problem. If you still are having these issues, try relaxing your memory timings.

[groovin]

my boxes are all extremely stable... only problem i had was using 2.6.8 kernel... for some reason it liked to stall out during bootup. after i switched to >2.6.8 everything worked great.

[SnEptUne]

I have been using Gentoo for 2 years. But it seems like my system are just getting more and more problems, such as obsidian+ segfault on me for no reason after I upgraded glibc, samba doesn't work anymore, etc...

I don't do emerge -uDp world at all because it is known to break system. I use the default CFLAGS, doesn't use ~x86 keywords except on tetex and other user applications. But there are always suprise for me.

Most of my applications are very out of date because I don't want to risk updating any package unless it is really needed. But when I do install new package because of there are vulernability according to GLSA, and it upgrades other system library, such as baselayout, it is a surprise for me when my network no longer work because they changed the init.d script (Since when am I suppose to use symlink for net.eth0? It sure wasn't mentioned in 2004 handbook).

So my question is this: Is Gentoo really stable at all? Can I use it on a mission critical server? It must be stable and available all the time. Even a minute down time is not acceptable.
_________________
"There will be more joy in heaven over the tear-bathed face of a repentant sinner than over the white robes of a hundred just men." (LM, 114)

[renrutal]

Yeah, you could use it as a mission critic system, but only if you treat it like any other mission critic system: test test test then deployment.

It's not like the developers of the other distros don't get the same errors we do. Of course they do. Just because Gentoo passes it to the users, it doesn't mean the distro is unstable, it just shows how great the upstream's programming is. I'm also sure there are errors caused by the Gentoo way of manage the packages, but I'm sure in all these years they filtered a lot of evil errors. But sometimes, some of them get through.

[Daemonax]

Gentoo is one of the more solid distros around. Lots of people seem to swear by Debian for stability though.
_________________
The God idea is growing more impersonal and nebulous in proportion as the human mind is learning to understand natural phenomena and in the degree that science progressively correlates human and social events. -- Emma Goldman

[effloresce]

[Comfortably_Numb]

[wswartzendruber]

[petrjanda]

[wswartzendruber]

[-=GGW=- $ol!d $n4>|e]