lubuntu-qa team mailing list archive

[JM <meets@xxxxxx>]

On Fri, 11 Oct 2013 10:47:21 +0200
Nio Wiklund <nio.wiklund@xxxxxxxxx> wrote:

> [Results after installation at the end]
> 
> -------- Original Message --------
> Subject: final Lubuntu i386 desktop live
> Date: Fri, 11 Oct 2013 09:06:30 +0200
> From: Nio Wiklund <nio.wiklund@xxxxxxxxx>
> To: lubuntu-qa@xxxxxxxxxxxxxxxxxxx <lubuntu-qa@xxxxxxxxxxxxxxxxxxx>
> 
> Hi!
> 
> I have started testing the final Lubuntu i386 desktop live session, and
> found the bugs 940919 and 1213837.
> 
> Slow but no freeze with low RAM
> 
> I limited RAM to 256 MB, and it worked. no freeze while I installed htop
> at the same time as I opened Firefox. I could run htop and watch
> swapping. Firefox was extremely slow, and you hardly browse to
> ubuntu.com (I saw the top of the page but could not scroll, even after
> waiting for minutes). So live session with 256 MB is no go with this
> version in this computer
> 
> http://www.toshiba.se/laptops/satellite-pro/c850/satellite-pro-c850-19w/
> 
> if you want to browse the internet. I could close Firefox and see the
> RAM being released.
> 
> Maybe old hardware and an installed system eats less RAM, so it might
> work better with low RAM, but the main reason of this test was to
> stress-test zRAM.
> 
> Best regards
> Nio
> ------------------------
> Hi again,
> 
> Installation worked with 512 MB RAM. Not the same freeze as with beta2,
> but the restart after installation was not clean. It froze with only
> back-light on, and I had to shut off with the button. After booting the
> installed system worked as it should.
> 
> Firefox offered to install flash 'Plugin Finder Service 2', but it got
> stuck forever (a bug?). But sudo apt-get flashplugin-installer works and
> lets me watch youtube videos. Firefox works reasonably well as installed
> with 256 MB RAM (much better than live in the same computer).
> 
> I did not identify any of the old bugs.
> 
> Best regards
> Nio

Hi,

I would have liked to read from your tests if the machine slows down by swapping too early
while some memory is still available or if it uses most of the physical RAM before doing
so. You could see it by having htop running in a console and keep on eye on proc and ram
use at the top of the console while starting and using applications.

I have done tests with ZRAM, in an iso I am doing (a personal remix done with
ubuntu-builder) and have not finished yet. It consists in once having the configuration
default provided by the zram-config package, then in another ISO switch to the
configuration I have described before. 

I still have one test to do which could make a difference.

Linux kernels have been known to swap to disk too early since many years, and the
following configuration is a mean to limit the too early swappiness.
 

So the test I want to do at last will consist in using the default zram-config
configuration and add just this: 


**********
vm.swappiness=0 
vm.vfs_cache_pressure=50 

# Uncomment the next line if we are running a laptop 
# vm.laptop_mode = 1

**********

as a file such as 50-local.conf in /etc/sysctl.d.

I wanted to try in Lubuntu but ubuntu-builder has not been able to redo a bootable ISO
from from the build directory after I added the file in the Lubuntu filesystem and
generated a new ISO.

If you have some means of testing this setup in your machine it could be interesting. (As
live I fancy it could be done with an install to USB with USB Creator and then having a
persistant mode, add the file, and reboot to the persistant mode once more to test the
setup, in your install you could just add the setup and see while using and with htop if
that makes a difference after some time using Firefox or else?)

What I don't know yet either is if having zram module loaded and at work increases the
linux kernel swappiness tendency or if it is totally unrelated : I had never wondered
about it so far.

Regards,
Mélodie