[QA] Release Meeting 2011-12-09

[Patrick Wright]

On Mon, Dec 12, 2011 at 3:49 PM, Steve Langasek
<steve.langasek at canonical.com> wrote:
> Hi Patrick,
>
> On Mon, Dec 12, 2011 at 08:16:16AM -0700, Patrick Wright wrote:
>> >> On Fri, Dec 09, 2011 at 09:34:43AM +0100, Jean-Baptiste Lallement wrote:
>
>> >>>  * Boot Speed testing
>> >>>    * Daily boot speed testing run on Desktop i386
>> >>>    * Report: http://reports.qa.ubuntu.com/reports/boot-speed/
>
>> >> Today's test for the Dell Vostro shows that ureadahead is not preloading
>> >> files as intended:
>
>> >> http://reports.qa.ubuntu.com/reports/boot-speed/dell-vostro-3400/2011-12-09_09-59-00/bootchart.png
>
>> >> This could be due to either a bug in ureadahead or a bug in the test
>> >> configuration.  (It does not appear to be caused by using an SSD; looking
>> >> at
>> >> the device name listed in dmesg, this seems to be a rotational disk model.
>> >> It would probably be helpful if this information were easily visible in
>> >> some
>> >> sort of summary of the machine information.)
>> If the info you require isn't listed in lspci file please provide me
>> with what specific data would be helpful to add. Providing the command
>> with options would be most helpful to me.
>
> Presumably you have information about the machines to know if the disks are
> rotational or SSD, but it's non-trivial to extract it from the system logs,
> which is why it would be useful to have it listed directly in the report
> somewhere.  It certainly wouldn't be listed in lspci (the disk is not a pci
> endpoint); the disk model is listed in the dmesg output, but it requires a
> google search from there to figure out what that disk is.  If there's a way
> to automatically tell from the kernel if a disk is an SSD, I don't know it.
I can add this information to test_info.


>
>> >> Is the methodology used for boot speed testing described somewhere?  I'd
>> >> like to try to reproduce this and figure out what's happening.  Not having
>> >> ureadahead working correctly causes a huge amount of noise in boot charts;
>> >> for instance, today's chart shows lightdm hitting the disk for a full
>> >> second
>> >> before it gets around to starting X (which gets counted against the
>> >> plumbing
>> >> budget), and this almost entirely accounts for the difference in plumbing
>> >> time in yesterday's run vs. today's on this machine.  There's no way to
>> >> tell from the currently available info if this is a regression in lightdm,
>> >> or just noise as a result of ureadahead not running correctly.
>
>> Automation aside, its really simple if you want to do this manually.
>> 1. Get the latest daily iso (desktop or alternate)
>> 2. Install from iso
>> 3. Install bootchart package on startup
>> 4. Reboot again
>> 5. Check bootchart results
>
> Thanks, having this spelled out is useful, because there's a bug in this
> methodology. ;)  You must reboot *twice* after installing the bootchart
> package, because the act of installing bootchart invalidates the ureadahead
> packs on the filesystem (/var/lib/dpkg/info/ureadahead.triggers, .postinst).
> So the first reboot after installing the bootchart package is a ureadahead
> profiling run with ureadahead not actually doing anything to help boot
> speed; then on a second reboot, ureadahead has a pack available that it can
> use for measuring.
>
> Please reboot twice after installing bootchart, as part of all the boot
> speed tests.
I will make sure this is done from now on.

>
>> You can add debugging to dmesg if that will provide information useful to
>> you.
>
> Well, given that my primary interest is the "Plumbing" timing, dmesg is not
> necessarily the most useful log for me :)
>
> Cheers,
> --
> Steve Langasek                   Give me a lever long enough and a Free OS
> Debian Developer                   to set it on, and I can move the world.
> Ubuntu Developer                                    http://www.debian.org/
> slangasek at ubuntu.com                                     vorlon at debian.org