[gentoo-user] ATLAS@home and VirtualBox

[gentoo-user] ATLAS@home and VirtualBox

[Peter Humphrey]

Hello list,

Has anyone here succeeded in getting this BOINC project to run? No matter 
what I try, I get "task postponed 86400.000000 sec: Communication with VM 
Hypervisor failed."

I've tried all available versions of gentoo-sources, virtualbox and boinc to 
no avail. Virtualization and Intel Vt-D are switched on in the UEFI BIOS. I 
can't see anything else in there that might affect this.

In case it's a GPU problem I've also tried all available versions of amdgpu-
ucode and linux-firmware. This is a new box with a Radeon R9 380 display card 
which shows up thus:

# lspci -ns 01:00.0
01:00.0 0300: 1002:6938 (rev f1)

One straw to clutch at is that the one VirtualBox client I run is 32-bit. Is 
there a knack to getting 64-bit virtualisation to work?

(I won't ask why 11 digits of precision are needed to represent 1 day.)

-- 
Rgds
Peter

Re: [gentoo-user] ATLAS@home and VirtualBox

[Gregory Woodbury]

[-- Attachment #1: Type: text/plain, Size: 750 bytes --]

Is the "user" boinc runs as in the vboxusers" group?
By default, boinc runs as a non-root (daemon?) user and if its group is not
in the vboxusers group,
VirtualBox will not start or run for the user.

The recommended workaround is to run boinc as a real user (generally "you")
so
that things operate properly.  Additionally, the actual way that the cernvm
runs can change "just a bit" and it will upset the balance of getting VBox
to run
boinc tasks.

On Tue, Apr 26, 2016 at 5:40 AM, Peter Humphrey <peter@prh.myzen.co.uk>
wrote:

> Has anyone here succeeded in getting this BOINC project to run? No matter
> what I try, I get "task postponed 86400.000000 sec: Communication with VM
> Hypervisor failed."
>


-- 
G.Wolfe Woodbury
redwolfe@gmail.com

[-- Attachment #2: Type: text/html, Size: 1377 bytes --]

Re: [gentoo-user] ATLAS@home and VirtualBox

[Peter Humphrey]

On Tuesday 26 Apr 2016 19:54:09 Gregory Woodbury wrote:
> On Tue, Apr 26, 2016 at 5:40 AM, Peter Humphrey <peter@prh.myzen.co.uk>
> wrote:
> > Has anyone here succeeded in getting this BOINC project to run? No
> > matter
> > what I try, I get "task postponed 86400.000000 sec: Communication with
> > VM
> > Hypervisor failed."
> Is the "user" boinc runs as in the vboxusers" group?
> By default, boinc runs as a non-root (daemon?) user and if its group is
> not in the vboxusers group,
> VirtualBox will not start or run for the user.
> 
> The recommended workaround is to run boinc as a real user (generally
> "you") so that things operate properly.  Additionally, the actual way that
> the cernvm runs can change "just a bit" and it will upset the balance of
> getting VBox to run boinc tasks.

Thanks for the reply. Yes, boinc is running as me, as shown below, and I am
in the vboxusers group. Boinc is also in that group. All the files and
directories under ~/boinc and ~/.VirtualBox are owned prh:prh. But I've just
noticed that the files under ~/.VirtualBox are all 0600; I wonder if that's
thet problem: that the boinc user can't read them. But then boinc is running
as me, the owner of all the files involved.

I'm going round in circles again.

$ ps U prh 
  710 ?        SNsl   4:10 /usr/bin/boinc_client --daemon --dir /home/prh/boinc --redirectio
  880 ?        Sl    22:23 /usr/bin/boincmgr
[...]
19613 ?        RN   372:47 ../../projects/www.cosmologyathome.org/camb_2.16_i686-pc-linux-gnu params.ini
[...]
24897 ?        SNl    0:08 ../../projects/pogs.theskynet.org_pogs/wrapper_x86_64-pc-linux-gnu_4.02
[...]
28001 ?        RN     7:17 ./fit_sed 10 filters.dat observations.dat
[...]

-- 
Rgds
Peter

Re: [gentoo-user] ATLAS@home and VirtualBox

[Peter Humphrey]

On Wednesday 27 Apr 2016 10:03:58 I wrote:

--->8

> I've just noticed that the files under ~/.VirtualBox are all 0600; I wonder
> if that's thet problem: that the boinc user can't read them. But then
> boinc is running as me, the owner of all the files involved.

Nope. After setting files to 0644 I see no change in behaviour.

-- 
Rgds
Peter

SOLVED Re: [gentoo-user] ATLAS@home and VirtualBox

[Peter Humphrey]

On Wednesday 27 Apr 2016 10:03:58 Peter Humphrey wrote:
> On Tuesday 26 Apr 2016 19:54:09 Gregory Woodbury wrote:
> > On Tue, Apr 26, 2016 at 5:40 AM, Peter Humphrey <peter@prh.myzen.co.uk>
> > 
> > wrote:
> > > Has anyone here succeeded in getting this BOINC project to run? No
> > > matter
> > > what I try, I get "task postponed 86400.000000 sec: Communication with
> > > VM
> > > Hypervisor failed."
> > 
> > Is the "user" boinc runs as in the vboxusers" group?
> > By default, boinc runs as a non-root (daemon?) user and if its group is
> > not in the vboxusers group,
> > VirtualBox will not start or run for the user.
> > 
> > The recommended workaround is to run boinc as a real user (generally
> > "you") so that things operate properly.  Additionally, the actual way
> > that the cernvm runs can change "just a bit" and it will upset the
> > balance of getting VBox to run boinc tasks.
> 
> Thanks for the reply. Yes, boinc is running as me, as shown below, and I
> am in the vboxusers group. Boinc is also in that group. All the files and
> directories under ~/boinc and ~/.VirtualBox are owned prh:prh. But I've
> just noticed that the files under ~/.VirtualBox are all 0600; I wonder if
> that's thet problem: that the boinc user can't read them. But then boinc
> is running as me, the owner of all the files involved.
> 
> I'm going round in circles again.
> 
> $ ps U prh
>   710 ?        SNsl   4:10 /usr/bin/boinc_client --daemon --dir
> /home/prh/boinc --redirectio 880 ?        Sl    22:23 /usr/bin/boincmgr
> [...]
> 19613 ?        RN   372:47
> ../../projects/www.cosmologyathome.org/camb_2.16_i686-pc-linux-gnu
> params.ini [...]
> 24897 ?        SNl    0:08
> ../../projects/pogs.theskynet.org_pogs/wrapper_x86_64-pc-linux-gnu_4.02
> [...]
> 28001 ?        RN     7:17 ./fit_sed 10 filters.dat observations.dat
> [...]

The fix has already been done, silently - no outward evidence other than the 
dates on a couple of files. I only found out when I compared a new 
installation with what I'd backed up a few hours earlier.

The problem was that /etc/init.d/boinc was starting the client with an su 
command, but now a start-stop-daemon call has been substituted. Job done. I 
can now move on to the next problem.  :)

-- 
Rgds
Peter

Re: SOLVED Re: [gentoo-user] ATLAS@home and VirtualBox

[Peter Humphrey]

On Monday 02 May 2016 11:16:28 I wrote:

> The fix has already been done, silently - no outward evidence other than
> the dates on a couple of files. I only found out when I compared a new
> installation with what I'd backed up a few hours earlier.

It seems there was a change-log entry, but I didn't see it. This is what I
saw:

$ equery c boinc
*boinc-7.4.42-r1 (09 Sep 2015)

  09 Sep 2015; Marius Brehler <marbre@linux.sungazer.de> -boinc-7.4.42.ebuild,
  +boinc-7.4.42-r1.ebuild:
  Switch to wxGTK 3.0, fixes bug #556670

  Package-Manager: portage-2.2.20.1

  24 Jan 2016; Michał Górny <mgorny@gentoo.org> metadata.xml:
  Replace all herds with appropriate projects (GLEP 67)

  Replace all uses of herd with appropriate project maintainers, or no
  maintainers in case of herds requested to be disbanded.

  24 Jan 2016; Michał Górny <mgorny@gentoo.org> metadata.xml:
  Set appropriate maintainer types in metadata.xml (GLEP 67)

What is the approved method of reading change-logs nowadays?

-- 
Rgds
Peter

Re: SOLVED Re: [gentoo-user] ATLAS@home and VirtualBox

[Andrew Savchenko]

[-- Attachment #1: Type: text/plain, Size: 1797 bytes --]

Hi,

On Tue, 03 May 2016 09:56:29 +0100 Peter Humphrey wrote:
> On Monday 02 May 2016 11:16:28 I wrote:
> 
> > The fix has already been done, silently - no outward evidence other than
> > the dates on a couple of files. I only found out when I compared a new
> > installation with what I'd backed up a few hours earlier.
> 
> It seems there was a change-log entry, but I didn't see it. This is what I
> saw:
> 
> $ equery c boinc
> *boinc-7.4.42-r1 (09 Sep 2015)
> 
>   09 Sep 2015; Marius Brehler <marbre@linux.sungazer.de> -boinc-7.4.42.ebuild,
>   +boinc-7.4.42-r1.ebuild:
>   Switch to wxGTK 3.0, fixes bug #556670
> 
>   Package-Manager: portage-2.2.20.1
> 
>   24 Jan 2016; Michał Górny <mgorny@gentoo.org> metadata.xml:
>   Replace all herds with appropriate projects (GLEP 67)
> 
>   Replace all uses of herd with appropriate project maintainers, or no
>   maintainers in case of herds requested to be disbanded.
> 
>   24 Jan 2016; Michał Górny <mgorny@gentoo.org> metadata.xml:
>   Set appropriate maintainer types in metadata.xml (GLEP 67)
> 
> What is the approved method of reading change-logs nowadays?

If you are using git sync, just go to /usr/portage/sci-misc/boinc
and run `git log .` there.

If you are using rsync sync, read end of the ChangeLog file, tail
-n 50 should do the job. 'equery c boinc' should work too.

Looks like your problem comes from broken rsync mirror with outdated
ChangeLog files. Also take into account that ChangeLogs are now
generated from git logs and this takes time, usually something
about an hour (maybe half an hour, I'm not sure).

Just tested on one of my systems using rsync, 'equery c boinc'
works fine. Make sure your app-portage/gentoolkit is up to date.

Best regards,
Andrew Savchenko

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: SOLVED Re: [gentoo-user] ATLAS@home and VirtualBox

[Peter Humphrey]

On Tuesday 03 May 2016 18:47:39 Andrew Savchenko wrote:
> On Tue, 03 May 2016 09:56:29 +0100 Peter Humphrey wrote:

--->8

> > What is the approved method of reading change-logs nowadays?
> 
> If you are using git sync, just go to /usr/portage/sci-misc/boinc
> and run `git log .` there.
> 
> If you are using rsync sync, read end of the ChangeLog file, tail
> -n 50 should do the job. 'equery c boinc' should work too.

Okay, thanks.

> Looks like your problem comes from broken rsync mirror with outdated
> ChangeLog files.

No, I see it now: 'equery c boinc' returns the ChangeLog section that refers 
to the currently installed version, not to any earlier or later ones. If I 
know what later versions are available I can do something like "equery c 
=sci-misc/boinc-7.6.31" but I hadn't done so until just now.

What's happened, I think, is that the init script change introduced into the 
latest version has been silently back-ported to the previous version. That's 
why I didn't see it in the ChangeLog, and it explains why I didn't get it 
during routine updates - only when I uninstalled the package and reinstalled 
it.

A learning point has emerged here: it's dangerous to specify a version in 
package.keywords.

I can't say I'm delighted to have spent three weeks in fruitless bug 
chasing, just because of an undocumented change that had already fixed my 
problem behind my back. At least I understand what's going on now. Thanks 
for your help, Andrew.

-- 
Rgds
Peter