Random crashes of LTSP client

[karaluh]

Every couple of minutes of regular work connection crashes and client is thrown to the login screen. It crashes only when there's some activity, I left the client loged on overnight and the crash didn't occur. The crash may be somewhat related to increased IO/CPU/video activity, it is 100% reproducible when starting Firefox session with > 10 tabs. The client is some old Compaq Deskpro 766 MHz 128 MB RAM, PXE booted. Server is on a VM.

ProblemType: Bug
Architecture: amd64
Date: Thu Jan 28 10:34:27 2010
DistroRelease: Ubuntu 9.10
Package: ltsp-server-standalone 5.1.90-0ubuntu3
PackageArchitecture: all
ProcEnviron:
 LANGUAGE=
 LANG=pl_PL.UTF-8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.31-18.55-server
SourcePackage: ltsp
Uname: Linux 2.6.31-18-server x86_64
XsessionErrors:
 (npviewer.bin:13381): Gtk-WARNING **: /usr/lib/gtk-2.0/2.10.0/engines/libqtcurve.so: wrong ELF class: ELFCLASS64
 (npviewer.bin:14154): Gtk-WARNING **: /usr/lib/gtk-2.0/2.10.0/engines/libqtcurve.so: wrong ELF class: ELFCLASS64
 (npviewer.bin:14397): Gtk-WARNING **: /usr/lib/gtk-2.0/2.10.0/engines/libqtcurve.so: wrong ELF class: ELFCLASS64
 (npviewer.bin:15242): Gtk-WARNING **: /usr/lib/gtk-2.0/2.10.0/engines/libqtcurve.so: wrong ELF class: ELFCLASS64

[Scott Balneaves (sbalneav)]

Do you have NBD_SWAP enabbled? In all likelihood, your thin client it running out of ram.

Firefox allocates pixmap cache in the X server. This causes the X server to consume more memory. When the kernel runs out of ram, if it has no swap defined, it goes on a killing spree to try and stay alive, killing the process that's consuming the most ram.

Which, on a thin client, is X.

[karaluh (karaluh)]

> Do you have NBD_SWAP enabbled?

No, all config files except dhcpd.are untouched by me.

[karaluh (karaluh)]

I forgot, the swap partition size is 1GB

[Alkis Georgopoulos (alkisg)]

NBD_SWAPping is broken in Karmic, so you'd need a very very recent LTSP version to see if swap helps:
http://bazaar.launchpad.net/~ltsp-upstream/ltsp/ltsp-trunk/revision/1678

The swap partition on your server is irrelevant.
If you have a swap partition *on the clients* (i.e. if they have a hard disk) you can enable it by putting USE_LOCAL_SWAP=True in lts.conf.

[karaluh (karaluh)]

I'll retest in Lucid.

[Alkis Georgopoulos (alkisg)]

The fix isn't published on Lucid (yet?). I think NBD swap worked in Hardy, though, so if you have a CD around you could test with Hardy.

One way to test with Lucid would be to add the following repository which has very recent LTSP versions:

1) Install Lucid

2) sudo add-apt-repository ppa:ts.sch.gr
    sudo apt-get update
    sudo apt-get dist-upgrade

3) sudo chroot /opt/ltsp/i386 add-apt-repository ppa:ts.sch.gr
    sudo chroot /opt/ltsp/i386 apt-get update
    sudo chroot /opt/ltsp/i386 apt-get dist-upgrade
    sudo chroot /opt/ltsp/i386 update-initramfs -u
    sudo ltsp-update-kernels
    sudo ltsp-update-image

[Christian (c-pradelli)]

Running latest Ubuntu Lucid I'm having a similar problem in all my machines that have old intel motherboard whith 8MB video onboard, all the machines have local swap enabled.
Client randomly crash and go to login.

[Marcos Saraiva (msaraiva)]

I can reproduce this same situation every time some of our users run Firefox. When they try to log on to our corporate intranet, everytime the user presses "enter" to login, X crashes and ldm restarts.

[Christian (c-pradelli)]

Marcos, try to use opera to login into the intranet.

In my case I need to make 2 fixes:

1) in intel video machines, disabling acpi in client kernel solves most of my problems.

2) other crash happens when firefox opening some web pages with flash, here I use opera.

[Marcos Saraiva (msaraiva)]

This bug is getting weirder. It's not only on Firefox that triggers a the crash, but even using OpenOffice. The user just have to press "enter", and then the session crashes.

[Alkis Georgopoulos (alkisg)]

Marcos, if your clients hang whenever you press [Enter], please file a separate bug. It could e.g. be something related to plymouth, like this one: https://bugs.launchpad.net/ubuntu/+source/plymouth/+bug/516412

[Christian (c-pradelli)]

Marcos, could you please post the hardware configuration of the client?

[Launchpad Janitor (janitor)]

[Expired for ltsp (Ubuntu) because there has been no activity for 60 days.]