OpenVZ Forum


Home » Mailing lists » Devel » Catching the console
Catching the console [message #24940] Wed, 12 December 2007 11:27 Go to next message
Enrico Weigelt is currently offline  Enrico Weigelt
Messages: 31
Registered: July 2006
Member
Hi folks,

by an recent post @users list I learned, I'm not the only one who
likes to catch the console at VE bootup. There are several ways
to work around (within the VE), but this actually doesn't *solve*
the problem.

No I'd like to request an feature for catching the console.
IMHO, it's enough to have an option to vzctl start, which tells
it to not to detach and bring the VE's console to the current tty.

How could this be done ?


cu
-- 
---------------------------------------------------------------------
 Enrico Weigelt    ==   metux IT service - http://www.metux.de/
---------------------------------------------------------------------
 Please visit the OpenSource QM Taskforce:
 	http://wiki.metux.de/public/OpenSource_QM_Taskforce
 Patches / Fixes for a lot dozens of packages in dozens of versions:
	http://patches.metux.de/
---------------------------------------------------------------------
Re: Catching the console [message #25024 is a reply to message #24940] Thu, 13 December 2007 08:16 Go to previous messageGo to next message
rkagan is currently offline  rkagan
Messages: 11
Registered: May 2006
Location: Moscow
Junior Member
On Wed, Dec 12, 2007 at 12:27:01PM +0100, Enrico Weigelt wrote:
> No I'd like to request an feature for catching the console.
> IMHO, it's enough to have an option to vzctl start, which tells
> it to not to detach and bring the VE's console to the current tty.

Unfortunately this is much trickier than that.  The problem is that at
some point you'll want to detach the VE0's end; with standard Unix98
ptys this will result in SIGHUP being sent to all processes in the VE
whose controlling terminal is the console.  Besides, there are certain
implications for online migration of the VEs - you should leave no
references to the VE0 objects that can't be reconstructed on the
destination node.

So it looks like there's no pure userspace solution.  OTOH adding the
kernel support for it doesn't look impossible off-hand; I'll have a
look.

Roman.
RE: Catching the console [message #25025 is a reply to message #25024] Thu, 13 December 2007 08:25 Go to previous messageGo to next message
Dietmar Maurer is currently offline  Dietmar Maurer
Messages: 52
Registered: March 2007
Member
Hi all,

here is a patch for vzctl to catch init output. output is written to VEROOT/var/log/init.log

I tested it with some debian templates (sysvinit), but it would be great is someone else tries/test it with other templates.

- Dietmar 

> -----Ursprüngliche Nachricht-----
> Von: devel-bounces@openvz.org 
> [mailto:devel-bounces@openvz.org] Im Auftrag von Roman Kagan
> Gesendet: Donnerstag, 13. Dezember 2007 09:16
> An: Enrico Weigelt
> Cc: OpenVZ developer list
> Betreff: Re: [Devel] Catching the console
> 
> On Wed, Dec 12, 2007 at 12:27:01PM +0100, Enrico Weigelt wrote:
> > No I'd like to request an feature for catching the console.
> > IMHO, it's enough to have an option to vzctl start, which 
> tells it to 
> > not to detach and bring the VE's console to the current tty.
> 
> Unfortunately this is much trickier than that.  The problem 
> is that at some point you'll want to detach the VE0's end; 
> with standard Unix98 ptys this will result in SIGHUP being 
> sent to all processes in the VE whose controlling terminal is 
> the console.  Besides, there are certain implications for 
> online migration of the VEs - you should leave no references 
> to the VE0 objects that can't be reconstructed on the 
> destination node.
> 
> So it looks like there's no pure userspace solution.  OTOH 
> adding the kernel support for it doesn't look impossible 
> off-hand; I'll have a look.
> 
> Roman.
>
  • Attachment: env.c.diff
    (Size: 2.64KB, Downloaded 212 times)
Re: Catching the console [message #25398 is a reply to message #25024] Sat, 22 December 2007 10:36 Go to previous message
Enrico Weigelt is currently offline  Enrico Weigelt
Messages: 31
Registered: July 2006
Member
* Roman Kagan <rkagan@sw.ru> wrote:

Hi,

> Unfortunately this is much trickier than that.  The problem is that at
> some point you'll want to detach the VE0's end; with standard Unix98
> ptys this will result in SIGHUP being sent to all processes in the VE
> whose controlling terminal is the console.  Besides, there are certain
> implications for online migration of the VEs - you should leave no
> references to the VE0 objects that can't be reconstructed on the
> destination node.

I just need it for debugging (eg. the startup process where no
remote access is available yet), so these problems aren't
important there.

If it really should be detachable, I suggest something like
screen(1), maybe implemented as 9P server.


cu
-- 
---------------------------------------------------------------------
 Enrico Weigelt    ==   metux IT service - http://www.metux.de/
---------------------------------------------------------------------
 Please visit the OpenSource QM Taskforce:
 	http://wiki.metux.de/public/OpenSource_QM_Taskforce
 Patches / Fixes for a lot dozens of packages in dozens of versions:
	http://patches.metux.de/
---------------------------------------------------------------------
Previous Topic: [PATCH net-2.6.25 3/3] Uninline the inet_twsk_put function
Next Topic: can we suggest a patch?
Goto Forum:
  


Current Time: Sun Sep 01 04:21:28 GMT 2024

Total time taken to generate the page: 0.05726 seconds