major problems with gentoo... EXTREAMLY unstable!

ee99ee2

Okay recently I've been freezeing my computer about once a day at least. Don't know why, not had time to look into it (been extreamly busy). I think it's WinMX though that's doing it (WinMX running on Win2k in VMware). I've got 768MB of RAM, so it's not like I'm running out of memory.

But anyway, last night was the last straw. I've got to get this problem fixed. I've tried once before to get things stable, and what did kinda worked. I just upgraded my video drivers.

But anyway, I'm afraid I've got much more serious problem on my hands. Last night when it froze right before I went to bed, I rebooted and it came back. But it was really weird. Alot of my config files in /etc were blown away. They were back to their defaults like they were on the CD when I installed my system 2 months ago. Here's just some of the files that were affected:

/etc/fstab
everything in /etc/conf.d/
/etc/hosts
/etc/hostname

I mentioned this in in #gentoo, and I was told that it's because my journal was out of date? I use XFS on all 3 main partitions (/, /var/mp3s, /home... /boot is ext2). I was also told in #gentoo that if I use reiserfs, it wouldn't happen. Someone said they deleted everything on their system (rm -rf /) on accident, and rebuilt it from the reiserfs journal? Something tells me that's not accurate, but that doesn't matter.

I want to get to the root of this problem and find out what's wrong and what the hell caused my config files to get blown away like that. /boot, swap, and / are all on one physical hard drive... /var/mp3s and /home is on another physical [identical] hard drive.

This all started about a month ago when I did an emerge -u world and portage was upgraded. I know that's what caused it, but that wouldn't cause my file system to bomb (or being to bomb like I think it is now). Is there anything I can do to try to tack it down?

Oh, one more note. I checked the logs. Absolutly nothing in them. A couple weeks ago, I turned on kernel debugging (Magic SysRq key), after all this starting happening, b/c I read when the kernel crashes, with that feature on it'll log more info to the logs about the crash. Dunno, but it's not helping...

Any ideas about this perplexing issue would be greatly appricated...

-ee99ee2
_________________
ServerMotion

BernieKe · Posted: Thu Feb 20, 2003 10:47 pm Post subject:

do you have a nvidia graphics card?
do you use the nvidia (or NVdriver) driver?

if yes: that's probably the cause, i'm suffering from the same problem (although no data loss after hard resets)
it seems that the nvdia-driver locks up X for some reason
i've found some solutions on the forums, but none of them worked

the only thing that DID work is using the old and trusted nv-driver (without glx extensions), but you won't be able to run quake3 (fe.) anymore

Bernard

SlCKB0Y · Posted: Fri Feb 21, 2003 3:11 am Post subject:

Do you have xscreensaver running? I found with some versions of nvidia drivers, and xscreensaver, after a particular screensaver had been on for a while, i would get a total lockup.

if you have them running, then try disabling them.

ee99ee2 · Posted: Fri Feb 21, 2003 3:32 am Post subject:

Xscreensaver does run... yes... but it rarely locks up when it's running... I'm not running one of the OpenGL screensavers... just a black screen

If I don't run xscreensaver, how else can I lock my screen so I have to enter a password to get back in?

Also, I tried downgradeng the nvidia kernel and glx driver (to 2960-r1). I've been as high as the 4xxx one. I had a freeze tonight about an hour after I installed the 2960 one... so that didn't help...

-ee99ee2
_________________
ServerMotion

Malakin · Posted: Fri Feb 21, 2003 4:30 am Post subject:

Try disabling agp and see what happens.

ee99ee2 · Posted: Fri Feb 21, 2003 4:33 am Post subject:

umm...... what

my card is an AGP card...

-ee99ee2
_________________
ServerMotion

BernieKe · Posted: Fri Feb 21, 2003 9:15 am Post subject:

i wasn't talking about an older nvidia-driver (I never tried it with the 2x series anyway (don't think they work on a geforce 4?)), i meant the standard nv you get whith X (like in putting nv instead of NVdriver or nvidia in the driver option, and disabling glx along with it)

Malakin · Posted: Fri Feb 21, 2003 7:27 pm Post subject:

ee99ee2 · Posted: Sun Feb 23, 2003 2:35 am Post subject:

Hmm.... I'll have to try that. Anyone know about my file system? Is there something I can do to check the validity of the journal??

-ee99ee2
_________________
ServerMotion

Bowyakka · Posted: Sun Feb 23, 2003 12:54 pm Post subject:

Lemme guess

You have XFS and prempt and nvidia drivers ??????

I found on my old mandrake box XFS to be rock solid (this was a machine without preempt) Even when it crashed I never lost files.

then I moved to gentoo and put XFS and premept in the same kernel , it seems (as far as I belive) that XFS utilises threads and that these threads get pre-empted when they should'nt be, I have had nothing but crashes all of which talk about xfs as being the cause. Also the crashes (like is mentioned in the gentoo install docs) seem to fry a lot of my data. Nvidia drivers seem to aggrivate this and increase the chances of these crashes .

My conclusion is to

1) use a kernel without preempt , lowlatency and the O1 patches( ie go and get a vanilla 2.4.20 and the apply the xfs patches found at oss.sgi.com to it)

2)Change filesystem to reiserfs (ext3 is too slow performance wise to bother with and jfs just doesn't have the feture set of xfs and reiser)

I changed my filesystems over (for my machine no mean feat it gave me an excuse to buy a dvd burner

) and used reiserfs , preempt and nvidia on the same kernel. After months of telling myself that it wasn't xfs it turns out that this was the cause of all my problems , I haven't had a major crash since ..........
_________________
glows in the dark

Lovechild · Posted: Sun Feb 23, 2003 3:58 pm Post subject:

Another good thing stabilitywise is to disallow optimzations on binutils, gcc and most importantly glibc - it will hurt your preformance a bit (but no worst than any other linux distro). I think the stablity is worth it.