Re: [E-devel] E2 update broken

[Carsten H. (T. R. <ra...@ra...>]

On Fri, 15 Jun 2012 23:19:50 +0200 Bertrand Jacquin <be...@me...> said:

> Hi in there,
> 
> As e3 was initially dedicated as a build host, I started installing some
> VM for that purpose that anyone can access easily. But I didn't document
> all this as not really terminated. A usage is more explicit :

e3/e4 have a big problem. disk space. :(

> $ ssh ssh@e3 help
> Usage: ssh {<opt>} <action>@e3.enlightenment.org <vm>
> 
> Available opt:
>   -t		for ssh
>   -X or -Y	for vnc/rdp
> 
> Available actions:
>   monitor
>   rdp
>   ssh
>   vnc
> 
> Available VM:
>   freebsd-i386
>   linux-amd64
>   linux-arm
>   linux-i386
>   linux-mips
>   linux-powerpc
>   winxp
> Connection to e3 closed.
> 
> So guys, you can do :
> 
>  # Access using SSH to freebsd-i386
>  $ ssh -t ssh@e3 freebsd-i386
> 
>  # Access to qemu monitor
>  $ ssh -t monitor@e3 freebsd-i386
> 
>  # Access to qemu VNC screen
>  $ ssh -t monitor@e3 freebsd-i386
> 
>  # Access using rdp to winxp
>  $ ssh -t rdp@e3 winxp
> 
> e3 is fixed since some minutes now, but all this is not (yet) really
> functional, this need some work.
> 
> What do you think about that ?
> 
> We need to able to clone VM if needed, build new one, stop, start, pause etc..
> 
> For pure compilation stuff, I would like to not use VM as it's not
> really performance killer. I also don't recommand to do that on main
> host as system become crap fast. Another solution is to maintain some
> chroot so host perf can be guarantee.

actually for pkg building i was originally going to make a bunch of chroots - on
e3/4 but disk space stopped me. then i just didnt have time to come back to it.

> What to you think ?
> 
> A proposal of needed chroot :
>  - linux-x86
>  - linux-x86_64
>  - linux-x32
> 
> About distro ? I'm familiar to gentoo that make better choice for long
> term purpose, but we can have chroot for many distro :
> 
>  - gentoo
>  - debian-lenny
>  - debian-squeeze
>  - RHEL5
>  - RHEL6
>  - Arch linux
>  - ...
> 
> A proposal of needed VM
> 
>  - freebsd7-i386
>  - freebsd7-amd64
>  - debian-arm6
>  - debian-arm7
>  - debian-mips
>  - debian-mips64
>  - debian-powerpc
>  - debian-powerpc64
> 
> Other things ?
> 
> D'ar gwener 15 a viz Mezheven 2012 e 02 eur 44, « Carsten Haitzler » he deus
> skrivet :
> > On Thu, 14 Jun 2012 22:45:35 +0200 Bertrand Jacquin <be...@me...>
> > said:
> > 
> > inc started an update and it was broken half0-way-through leaving webserver
> > barfing - i had to complete it and fix up some other things that broke -
> > like python modules.
> > 
> > the core problem is that we need an up to date system to build e and efl in
> > order to build docs...
> > 
> > i've been meaning to bring this up actually -beber... me and cedric were
> > throwing some ideas around and it may be time to replace e1/e2 machines.
> > they are now 6 years  old... and have been running without problems for
> > that time... hdd's included. <god i'm asking for trouble now!>. as part of
> > this we need a migration plan and of course setup of the new machines. here
> > is an idea/plan that came out of it:
> > 
> > 1. if we can get 2 new ones in side-by-side with the 2 old so we have a few
> > weeks of move-over time, if not, then bring down 1 (e1 to start) and set up
> > the new e2 replacement in its stead, then switch and swap out old e2 and
> > set up new box then. either way we can manage it as we have 2 rack slots
> > anyway.
> > 
> > 2. when we are done i think it'd be nice of us to donate the old machines to
> > osuosl to then share out between other oss projects as they deem fit. osuosl
> > have been good to us and we can be good back. :)
> > 
> > 3. now to new machines. i'm thinking this time around we might want to try
> > vm's. the vm space is mature now, so maybe a virtualbox "cluster" where the
> > host vm is really minimal and has nothing more than a few logins for admin
> > and ssh and everything else is stuffed into a vm. i'm a little hesitant as
> > i just detest layers... but hey - the question is going out to you as to
> > what you think. :)
> > 
> > 4. if we have vm's we should keep www/svn/git on 1, and create more vm's for
> > things like buildbot, document generation, etc. etc. - theses vm's can be
> > "bleeding edge".
> > 
> > 5. as they are vm's as long as we suspend, backup ram+fs image, and resume,
> > we can actually do much better backups AND split our vm's across the new
> > e1/e2 and migrate as needed. as we will be copying files from doc gen
> > machine to live webserver anyway - if they get copied from vm to vm on the
> > same box or across the network.. does it matter? sure - a bit in speed, but
> > hey.. same principle.
> > 
> > 6. if we have vms'.. we could get vincent to set up a windows vm. with
> > mingw32, sshd and so on set up so people can help out with acctually seeing
> > his windows build problems first hand as most of us baulk at the idea of
> > chasing down a copy of windows and a license to then install and figure out
> > mingw32/cygwin and friends jut to see a build error. :) of course such a vm
> > would have to be behind a machine-local nat so u have to bounce through
> > another vm to get to it...
> > 
> > 7. since we are talking windows vm's - we could maintain vm's for most major
> > linux distros, bsd's, solaris's and so on. again - put behind the nat. this
> > would expand our ability to have a buildbot farm and even build pkgs for
> > many os's.
> > 
> > p.s. the diea for now would be to get relatively beefy machines - xeon i7's
> > with 8 cores (4+4ht), 8 or 16gb ram, 2tb raid10 (1tb accessible) with maybe
> > a 128gb ssd for the host os and key files/data and maybe
> > websever/production vm's.
> > 
> > > Hi,
> > > 
> > > This is the game of an unstable install on a production system. dpkg is
> > > now OK, but we really should think a stable system.
> > > 
> > > Not sure the system will boot next time.
> > > 
> > > Beber
> > > 
> > > D'ar yaou 14 a viz Mezheven 2012 e 16 eur 08, « Jonas M. Gastal » he deus
> > > skrivet :
> > > > Hello everyone,
> > > > 
> > > > So it seems that someone recently(over the past 7 days) updated E2, and
> > > > in the process broke it. dpkg is not allowing any packages to be
> > > > installed because it complains the init system has got broken scripts.
> > > > This wouldn't be a big deal if not for the fact that the update
> > > > includes doxygen-1.8.1 which is bugged and produces horrific
> > > > documentation. I'm wondering if whoever updated it could help me fix
> > > > the init issues, I don't want to end up overwriting some custom script
> > > > someone might have created.
> > > 
> > > -- 
> > > Beber
> > 
> > 
> > -- 
> > ------------- Codito, ergo sum - "I code, therefore I am" --------------
> > The Rasterman (Carsten Haitzler)    ra...@ra...
> > 
> 
> -- 
> Beber


-- 
------------- Codito, ergo sum - "I code, therefore I am" --------------
The Rasterman (Carsten Haitzler)    ra...@ra...