12:50 p.m.
Weirdly the boot.iso fails to start X, but the kde live which is based on the same
boot.iso works.
ERR program: Error running metacity: SIGCHLD caught when trying to start the X server.
And fpaste X.log gets me:
Error Uploading: [Errno -2] Name or service not known
On another note, is there a way to get at least one full set (one boot.iso, one live
image, and one DVD image - if we're still going to have them) using a non-debug kernel
each week? Yesterday's rc6 kernel is a non-debug kernel, but the workstation image
failed to build today so I'm pretty sure that means the next attempt will use the then
current kernel which will end up being a debug kernel. They are so dreadfully slow… I just
won't use them. I'm sure many people who might be inclined to do more Rawhide
testing don't want to use debug kernels either because of how astronomically slow it
makes everything.
Chris Murphy
1:46 p.m.
On Fri, 2014-05-23 at 11:50 -0600, Chris Murphy wrote:
Weirdly the boot.iso fails to start X, but the kde live which is
based on the same boot.iso works.
lives aren't 'based on' boot.isos, they're separate compose processes.
ERR program: Error running metacity: SIGCHLD caught when trying to
start the X server.
the non-live installer uses metacity as its window manager. nothing else
does any more, so it's not surprising the live images (which all use
other window managers) don't have this issue.
And fpaste X.log gets me:
Error Uploading: [Errno -2] Name or service not known
odd, copy it off via a USB stick?
On another note, is there a way to get at least one full set (one
boot.iso, one live image, and one DVD image - if we're still going to
have them) using a non-debug kernel each week? Yesterday's rc6 kernel
is a non-debug kernel, but the workstation image failed to build today
so I'm pretty sure that means the next attempt will use the then
current kernel which will end up being a debug kernel. They are so
dreadfully slow… I just won't use them. I'm sure many people who might
be inclined to do more Rawhide testing don't want to use debug kernels
either because of how astronomically slow it makes everything.
I don't really find them 'astronomically' slow any more, now that
slub_debug is left off most of the time. Slightly slower than regular
kernels, sure, but perfectly fine for testing. Is it different on your
setup?
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net
5:43 p.m.
New subject: RC6 rawhide rocks
Today's RC6 rawhide installed via rsync and pxeboot is providing a
mostly pleasant experience.
For the last few RCx cycles, the first Rawhide 64 install has provided
a snappy, responsive desktop with Xfce. A day or two later the rawhide
install results in a sluggish desktop and a flurry of kernel OOPs.
on my office machine, a 3770-k, Asus p8z77-v le plus and GTX 670,
KDE startup stalls after displaying the hard drive icon. Gnome and
Xfce work normally.
KDE runs fine on an older machine.
YMMV.
--
Chuck Forsberg WA7KGX caf(a)omen.com www.omen.com
Developer of Industrial ZMODEM(Tm) for Embedded Applications
Omen Technology Inc "The High Reliability Software"
10255 NW Old Cornelius Pass Portland OR 97231 503-614-0430
11:56 p.m.
On May 23, 2014, at 12:46 PM, Adam Williamson <awilliam(a)redhat.com> wrote:
I don't really find them 'astronomically' slow any more, now that
slub_debug is left off most of the time. Slightly slower than regular
kernels, sure, but perfectly fine for testing. Is it different on your
setup?
kernel-core-3.15.0-0.rc6.git0.1.fc21.x86_64
# systemd-analyze
Startup finished in 905ms (kernel) + 1.445s (initrd) + 7.829s (userspace) = 10.180s
kernel-core-3.15.0-0.rc6.git1.1.fc21.x86_64
# systemd-analyze
Startup finished in 4.726s (kernel) + 14.750s (initrd) + 28.758s (userspace) = 48.235s
kernel-core-3.15.0-0.rc6.git1.1.fc21.x86_64 with slub_debug=-
# systemd-analyze
Startup finished in 4.666s (kernel) + 14.890s (initrd) + 28.142s (userspace) = 47.700s
And when the debug kernel is used for installing, the installs are likewise 4-5x longer.
For me, it's not usable. The VM is vastly slower, and its host gets hotter and runs
fans in noisy mode.
Chris Murphy
10:30 p.m.
On 05/27/2014 12:56 AM, Chris Murphy wrote:
On May 23, 2014, at 12:46 PM, Adam Williamson
<awilliam(a)redhat.com> wrote:
> I don't really find them 'astronomically' slow any more, now that
> slub_debug is left off most of the time. Slightly slower than regular
> kernels, sure, but perfectly fine for testing. Is it different on your
> setup?
kernel-core-3.15.0-0.rc6.git0.1.fc21.x86_64
# systemd-analyze
Startup finished in 905ms (kernel) + 1.445s (initrd) + 7.829s (userspace) = 10.180s
kernel-core-3.15.0-0.rc6.git1.1.fc21.x86_64
# systemd-analyze
Startup finished in 4.726s (kernel) + 14.750s (initrd) + 28.758s (userspace) = 48.235s
kernel-core-3.15.0-0.rc6.git1.1.fc21.x86_64 with slub_debug=-
# systemd-analyze
Startup finished in 4.666s (kernel) + 14.890s (initrd) + 28.142s (userspace) = 47.700s
And when the debug kernel is used for installing, the installs are likewise 4-5x longer.
For me, it's not usable. The VM is vastly slower, and its host gets hotter and runs
fans in noisy mode.
OK, would this "slow kernel" show itself as "long" pauses
between
packages when installing. Also, an install for a simple LXDE desktop
seems to be taking about 20 minutes. Is this a long or a good/short
time? All of this is withj a qemu-kvm virtual system with 2 VCPU, 2GB
memory and 8GB virtual disk. The input is a DVD iso on host disk.
Gene
11:42 p.m.
On May 27, 2014, at 9:30 PM, Gene Czarcinski <gczarcinski(a)gmail.com> wrote:
On 05/27/2014 12:56 AM, Chris Murphy wrote:
> On May 23, 2014, at 12:46 PM, Adam Williamson <awilliam(a)redhat.com> wrote:
>> I don't really find them 'astronomically' slow any more, now that
>> slub_debug is left off most of the time. Slightly slower than regular
>> kernels, sure, but perfectly fine for testing. Is it different on your
>> setup?
> kernel-core-3.15.0-0.rc6.git0.1.fc21.x86_64
> # systemd-analyze
> Startup finished in 905ms (kernel) + 1.445s (initrd) + 7.829s (userspace) = 10.180s
>
> kernel-core-3.15.0-0.rc6.git1.1.fc21.x86_64
> # systemd-analyze
> Startup finished in 4.726s (kernel) + 14.750s (initrd) + 28.758s (userspace) =
48.235s
>
> kernel-core-3.15.0-0.rc6.git1.1.fc21.x86_64 with slub_debug=-
> # systemd-analyze
> Startup finished in 4.666s (kernel) + 14.890s (initrd) + 28.142s (userspace) =
47.700s
>
> And when the debug kernel is used for installing, the installs are likewise 4-5x
longer. For me, it's not usable. The VM is vastly slower, and its host gets hotter and
runs fans in noisy mode.
>
OK, would this "slow kernel" show itself as "long" pauses between
packages when installing.
I think so yes.
Also, an install for a simple LXDE desktop seems to be taking about
20 minutes.
At least.
Is this a long or a good/short time?
It's long. Non-debug installs are much faster. I just did a live install with a
non-debug kernel that took 5 minutes.
Chris Murphy
3:01 p.m.
On 05/29/2014 12:42 AM, Chris Murphy wrote:
On May 27, 2014, at 9:30 PM, Gene Czarcinski
<gczarcinski(a)gmail.com> wrote:
> On 05/27/2014 12:56 AM, Chris Murphy wrote:
>> On May 23, 2014, at 12:46 PM, Adam Williamson <awilliam(a)redhat.com> wrote:
>>> I don't really find them 'astronomically' slow any more, now
that
>>> slub_debug is left off most of the time. Slightly slower than regular
>>> kernels, sure, but perfectly fine for testing. Is it different on your
>>> setup?
>> kernel-core-3.15.0-0.rc6.git0.1.fc21.x86_64
>> # systemd-analyze
>> Startup finished in 905ms (kernel) + 1.445s (initrd) + 7.829s (userspace) =
10.180s
>>
>> kernel-core-3.15.0-0.rc6.git1.1.fc21.x86_64
>> # systemd-analyze
>> Startup finished in 4.726s (kernel) + 14.750s (initrd) + 28.758s (userspace) =
48.235s
>>
>> kernel-core-3.15.0-0.rc6.git1.1.fc21.x86_64 with slub_debug=-
>> # systemd-analyze
>> Startup finished in 4.666s (kernel) + 14.890s (initrd) + 28.142s (userspace) =
47.700s
>>
>> And when the debug kernel is used for installing, the installs are likewise 4-5x
longer. For me, it's not usable. The VM is vastly slower, and its host gets hotter and
runs fans in noisy mode.
>>
> OK, would this "slow kernel" show itself as "long" pauses between
packages when installing.
I think so yes.
> Also, an install for a simple LXDE desktop seems to be taking about 20 minutes.
At least.
> Is this a long or a good/short time?
It's long. Non-debug installs are much faster. I just did a live install with a
non-debug kernel that took 5 minutes.
BTW, have you seen this problem:
https://bugzilla.redhat.com/show_bug.cgi?id=1101544
Basically, anaconda-yum goes off into the weeds (100% CPU) in the middle
of an install. I have a
strace -p <pid_of_anaconda-yum> >/mnt/xx/strace1.txt
There are also pauses which last a couple of minutes and then things
start up again. If you start strace after it is hung, it shows nothing.
Gene
11:04 p.m.
On May 23, 2014, at 12:46 PM, Adam Williamson <awilliam(a)redhat.com> wrote:
I don't really find them 'astronomically' slow any more,
now that
slub_debug is left off most of the time. Slightly slower than regular
kernels, sure, but perfectly fine for testing. Is it different on your
setup?
What about Rawhide media having two kernel options in its boot menu? The current
week's non-debug kernel, and then whatever the current debug kernel is?
Chris Murphy
11:01 a.m.
On Thu, 2014-05-29 at 22:04 -0600, Chris Murphy wrote:
On May 23, 2014, at 12:46 PM, Adam Williamson
<awilliam(a)redhat.com> wrote:
> I don't really find them 'astronomically' slow any more, now that
> slub_debug is left off most of the time. Slightly slower than regular
> kernels, sure, but perfectly fine for testing. Is it different on your
> setup?
What about Rawhide media having two kernel options in its boot menu?
The current week's non-debug kernel, and then whatever the current
debug kernel is?
What 'media' do you mean? The boot.iso? I didn't see that on the last
live I grabbed.
Obviously, it's a bug.
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net
1:02 p.m.
On 05/30/2014 09:01 AM, Adam Williamson wrote:
On Thu, 2014-05-29 at 22:04 -0600, Chris Murphy wrote:
> On May 23, 2014, at 12:46 PM, Adam Williamson <awilliam(a)redhat.com> wrote:
>> I don't really find them 'astronomically' slow any more, now that
>> slub_debug is left off most of the time. Slightly slower than regular
>> kernels, sure, but perfectly fine for testing. Is it different on your
>> setup?
>
> What about Rawhide media having two kernel options in its boot menu?
> The current week's non-debug kernel, and then whatever the current
> debug kernel is?
What 'media' do you mean? The boot.iso? I didn't see that on the last
live I grabbed.
My interpretation was that it was a suggestion, not an observation.
4:07 p.m.
On Fri, 2014-05-30 at 11:02 -0700, Samuel Sieb wrote:
On 05/30/2014 09:01 AM, Adam Williamson wrote:
> On Thu, 2014-05-29 at 22:04 -0600, Chris Murphy wrote:
>> On May 23, 2014, at 12:46 PM, Adam Williamson <awilliam(a)redhat.com>
wrote:
>>> I don't really find them 'astronomically' slow any more, now
that
>>> slub_debug is left off most of the time. Slightly slower than regular
>>> kernels, sure, but perfectly fine for testing. Is it different on your
>>> setup?
>>
>> What about Rawhide media having two kernel options in its boot menu?
>> The current week's non-debug kernel, and then whatever the current
>> debug kernel is?
>
> What 'media' do you mean? The boot.iso? I didn't see that on the last
> live I grabbed.
>
My interpretation was that it was a suggestion, not an observation.
oooh. yeah. that makes much more sense!
unfortunately I think it'd be a bit tricky to achieve.
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net
11:07 p.m.
On May 29, 2014, at 2:01 PM, Gene Czarcinski <gczarcinski(a)gmail.com> wrote:
>
BTW, have you seen this problem:
https://bugzilla.redhat.com/show_bug.cgi?id=1101544
I don't think so, so far all of my Rawhide installs complete (eventually).
Chris Murphy
3630
days inactive
3637
days old
11 comments
5 participants
participants (5)
-
Adam Williamson -
Chris Murphy -
Chuck Forsberg WA7KGX -
Gene Czarcinski -
Samuel Sieb