Using swap causes system to stop responding

djortz · 10-13-2005, 02:36 PM

[frustrated]
I didn't know it would be that hard to submit a bug to Suse. Do they want to keep their bugtracker a secret or what? I've been searching for an hour now (suselinux.com, google, novell) and the closest I got was getting an error after registring at bugzilla.novell.com.

Is there someone out there that has a clue where to submit my bug?

(Please don't tell me in 5 minutes, that would be bad for my head and monitor

)

Btw, the output of top before I did a hard reboot on my system:

Code:

top - 20:37:08 up  3:30,  5 users,  load average: 21.10, 10.53, 4.40
Tasks: 2835 total,  18 running, 2817 sleeping,   0 stopped,   0 zombie
Cpu(s):  2.6% us, 97.2% sy,  0.0% ni,  0.0% id,  0.0% wa,  0.0% hi,  0.2% si
Mem:   1024828k total,  1016540k used,     8288k free,     6752k buffers
Swap:  1028120k total,  1028120k used,        0k free,    12280k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
  171 root      16   0     0    0    0 R 24.3  0.0   1:07.25 kswapd0
 6394 root      15  -2  3632  332  320 D 20.4  0.0   0:24.99 hidd
 7437 sjors     17   0  191m 4168 2044 R 14.5  0.4   2:33.18 xmms
13514 sjors     17   0 39664 1756 1192 R 12.8  0.2   0:12.20 xmms-arts-helpe
 7436 sjors     16   0 85752 2132 1672 R  6.1  0.2   0:13.46 kmix
 7078 root      15   0 11444  580  512 S  4.8  0.1   0:09.11 powersaved
 7403 sjors     15   0 83392 2412 1756 S  3.5  0.2   0:06.48 kwin
13549 sjors     15   0  6340 2992  668 R  3.4  0.3   0:06.32 top
 7431 sjors     16   0 12592  608  476 R  2.8  0.1   0:05.50 ksysguardd
 7385 sjors     16   0  120m 2672 1696 S  2.2  0.3   0:11.06 kded
 7434 sjors     16   0  107m 2680 2172 S  1.7  0.3   0:11.40 knotify
 7399 sjors    -51   0 52108 2300 1380 S  1.2  0.2   2:52.70 artsd
 6569 root      17   0  570m  33m 1316 R  0.4  3.4  17:06.89 X
    1 root      16   0   724   56   32 S  0.0  0.0   0:01.01 init
    2 root      34  19     0    0    0 R  0.0  0.0   0:00.00 ksoftirqd/0
    3 root      10  -5     0    0    0 S  0.0  0.0   0:00.10 events/0
    4 root      11  -5     0    0    0 S  0.0  0.0   0:00.00 khelper

And free -m

Code:

do okt 13 20:35:33 CEST 2005
             total       used       free     shared    buffers     cached
Mem:          1000        992          8          0          6         12
-/+ buffers/cache:        973         27
Swap:         1004       1004          0

agtlewis · 10-13-2005, 06:10 PM

When your pc locks is your hard drive light lit up solid? I am having a similar problem when the system should be swapping, it just locks up. I am running fedora 4.

djortz · 10-13-2005, 06:15 PM

No, I don't think so. It isn't a very quiet disk, so I do hear it's activity. At lock up there is hardly any activity of the drive.

syg00 · 10-13-2005, 06:51 PM

djortz, there is a Suse forum here on LQ - maybe ask there about a bugzilla equivalent.

When you are in "top", try sorting by Virt - F (upper case) followed by o (oh - lowercase) should do it.
Am I the only one that thinks this looks a *little* high ???

Quote:

Tasks: 2835 total, 18 running, 2817 sleeping, 0 stopped, 0 zombie

- out of control fork/spawn maybe ???.

The high cpu% and no disk activity sounds like "wheel spinning" - chasing through control blocks looking for free slots to use. This gets very expensive when occupancy gets high - "high" in this case is probably 30%, although I have no numbers to confirm this.
Yet.

djortz · 10-14-2005, 01:26 AM

Quote:

djortz, there is a Suse forum here on LQ - maybe ask there about a bugzilla equivalent.

I did search there, but couldn't find it. I'll search again and then ask it there. Didn't want to open a new topic for such a little question

Quote:

Am I the only one that thinks this looks a *little* high ???

I did not look at that line, good point.
Top is filled with over 2000 lines containing more or less this:

Code:

14236 root      16   0  9008  948  920 S  0.0  0.1   0:00.00 bash        (10x)
14249 root      25   0  4628  324  324 S  0.0  0.0   0:00.00 grep        (5x)
(repeats and becomes)
17035 root      18   0  2652  440  360 R  0.0  0.0   0:00.00 grep
17042 root      18   0  2652  440  360 D  0.0  0.0   0:00.00 grep
17037 root      18   0  2652  440  360 D  0.0  0.0   0:00.00 grep
17043 root      18   0 12464 3408  948 R  0.0  0.3   0:00.00 bash
17044 root      18   0 12464 3392  932 R  0.0  0.3   0:00.00 bash

So Virt is very high.

Quote:

although I have no numbers to confirm this. Yet.

Just tell me what you want to know

Quote:

When you are in "top", try sorting by Virt

I suppose you now know what you want to know (the bash and grep processes)

agtlewis · 10-14-2005, 02:01 AM

Ok,

I know that grep is a search tool of some sort, but why in the world would it be using over 1000 different threads? Something must be in a loop, like a while loop that never ends or something? Is that similar to the problem?

I am a new linux user, what exactly is bash? I thought bash was something like a command interpreter behind the command prompt or something like that. Why would that be using so many threads?

[edit]
I noticed they are all running as root. Perhaps that can help you narrow down the problem since you aren't *supposed* to run many apps as root.
[/edit]

syg00 · 10-14-2005, 02:34 AM

Bash is the command shell - try a "man bash", or perhaps this will help.

Good point about root usage - djortz can it be run as a non-root user ???.
Try logging on (non-root) and issuing

Code:

ulimit -a

to check the number of processes limit.
Probably higher than what you are hitting, but you could try lowering it if you like and see what happens when the limit is reached.

djortz · 10-14-2005, 05:55 AM

I don't see a way to install packages as non-root.

With that in my ulimit -a won't be very useful I guess, but here it comes:

Code:

core file size        (blocks, -c) 0
data seg size         (kbytes, -d) unlimited
file size             (blocks, -f) unlimited
max locked memory     (kbytes, -l) 32
max memory size       (kbytes, -m) unlimited
open files                    (-n) 1024
pipe size          (512 bytes, -p) 8
stack size            (kbytes, -s) unlimited
cpu time             (seconds, -t) unlimited
max user processes            (-u) 8183
virtual memory        (kbytes, -v) unlimited

Again, the update/install should be run as root. Installing as a normal user could only be an option for testing purposes. I don't see any reason lowering this limits.

syg00 · 10-14-2005, 07:04 AM

Better talk to Novell then ...

djortz · 10-14-2005, 07:22 AM

Isn't that great, after logging in at bugzilla.novell.com (what seems to be the place to be) I get an error message. *puls out some hair*

I just tried to burn a DVD but k3b caussed the same problems when detecting the drives. I'll try to re-sign in at bugzilla, if that won't help, I'll have to re-install. I don't think that would help though.

Error at bugzilla:
Undefined subroutine &main::ValidateNewUser called at Bugzilla/Auth/Verify/iChain.pm line 204.
For help, please send mail to the webmaster (emailaddress), giving this error message and the time and date of the error.

*I did email the webmaster

Update:
Currently downloading Suse 10.0

foo_bar_foo · 10-14-2005, 03:10 PM

Quote:

Originally posted by djortz

I just tried to burn a DVD but k3b caussed the same problems when detecting the drives.

now that starts to sound like a kernel problem.
of worse yet a glibc problem.

might try just making yourself a non suse kernel and see if it works better

syg00 · 10-14-2005, 05:06 PM

Uggghh !!!.
Wonder if an incomplete update is screwing you. I'm with foo_bar_foo, a new kernel might be a good start.

But if the tool-chain is rooted, so are you. Especially on a "packaged" distro like that.

djortz · 10-14-2005, 05:10 PM

I just received an email from the novell webmaster. He'll take a look at it.

I finished downloading Suse 10.0, I think I will install that one since this is a ferly new system. New Suse version means new kernel, so it could solve things. There's only one problem: k3b doesn't work for me, so I have to burn the dvd from bash.

djortz · 10-15-2005, 09:57 AM

A little update..

Today I have backed up my /home and installed Suse10.0 over Suse9.3 (fresh install, formatted and everything). By now I have configured the system and it's now time to restore /home. I've done some tests and it seems OK. The system is less stressed (lower load, more memory free) so that is a good sign as well.

I would like to thank everybody that has put some effort in this topic. It's a pitty a need to reconfigure, but the problem is solved so I can trust my system again.

djortz · 10-16-2005, 05:31 AM

I've seen a little movie of computers in a shredder, a good place for mine....
After the upgrade to Suse10 everything looked OK, but no.. Today I tried to install lame (from rpm) and Yast gave up on me. This time it quited a little earlier with the same symptons, increasing memory usage, increasing load and lots of processes. Uninstalling the package did not work.
When I install in init3 there are no problems. It seems like my system can't cope with the stress.

Novell's webmaster seems to be not working on saturday or sunday, so I haven't been able to submit a bug.
So, problem not solved, stay tuned.

As a sidenote: I'm really pissed off right now, this things should only happen with M$ software

Edit
This is odd. I have added a new user and logged in as that user. Everything is OK with that user. I'll remove my first user account now and I'll config it step by step. It could be some program that is running that conflicts with being rooted. (Please let it be that) Perhaps there is a script in ~/bin/ that has the same name as another script (just a tought)

I can stop k3b taking down the system by killing X (ctrl+alt+backspace), haven't been able to test that for Yast