Xen 3.0.0 released!

Xen 3.0.0 released!

[Ian Pratt]

Folks,

Now seems a good time to call a Xen 3.0.0 release!  We've been seeing
good stability on the XenRT regression tests for the last couple of
weeks, and the number of bug reports submitted to bugzilla have
dropped right down. It's time to get a bigger group of people to start
beating up on it...

mercurial source: http://xenbits.xensource.com/xen-3.0-testing.hg

Along with the usual binary install tarball, we've created a new
live-iso demo CD, and some RPM packages for common linux
distros. Among the RPM downloads are 'xen-friendly' glibc packages.
See http://www.xensource.com/xen/downloads

[Please use the Torrent versions if you can; http downloads require
email registration so we can control server load over the next few
days.]

I expect we'll be spinning fixes to the 3.0 series fairly frequently
over the next few weeks. You'll be able to find the very latest
updates by pulling source from the 3.0-testing.hg tree, or using the
nightly builds. 

For the next few weeks, the xen-unstable.hg tree will be used to test
further releases in the 3.0.x series (we'll checkin the version of
linux 2.6.14 that has xen as a sub-arch that's currently sitting in
linux-2.6-xen.hg), before forking off to become the 3.1 development
branch once things have settled down on 3.0.

As usual, please send those bug reports to the mailing lists and
bugzilla!

Best,
Ian

Re: Xen 3.0.0 released!

[Michael Holzt]

> Now seems a good time to call a Xen 3.0.0 release!

The source code download tar.gz seems to disagree, as it contains an
'xen-unstable' directory. You probably want to change that :-)

Regards
Michael

-- 
      It's an insane world, but i'm proud to be a part of it. -- Bill Hicks

Re: [Xen-devel] Xen 3.0.0 released!

[Adam Heath]

On Mon, 5 Dec 2005, Ian Pratt wrote:

>
> Folks,
>
> Now seems a good time to call a Xen 3.0.0 release!  We've been seeing
> good stability on the XenRT regression tests for the last couple of
> weeks, and the number of bug reports submitted to bugzilla have
> dropped right down. It's time to get a bigger group of people to start
> beating up on it...

That's great.

I've had 3.0-unstable debs prepared for awhile now; hadn't uploaded them, for
several reasons:

1) xen-unstable required a 2.6.12 base kernel; this no longer exists in
   debian, so it wasn't possible for me to build-depend on it, so I could
   create a standard unified-diff.

   My fix for that, is to encapsulate the mkbuildtree logic into the
   kernel-package patch system, and not use a unified-diff at kernel compile
   time.  I've got the apply script done, but not the unpatch
   script(basically, the script looks into the tar, notes all modified and new
   files, saves them into debian/, and then during unpatch restores things).

2) vif-route.sh is buggy; it doesn't handle the case where multiple adapters
   exist.  So, dom0_ip(and the functions it calls) ends up returning 2 ip
   addresses, which makes ifconfig complain.

   I've got a fix hot-patched on our deployed server, I'll be sending that
   soon.

3) It appeared that when I tried the final 3.0.0 release, that xenstored was
   very sick.  I had a previous(20051116 snapshot) installed, and the tdb file
   was corrupted(sorta).  xenstore-list / shows to local paths, but no vm nor
   tools.  Removing the file 'fixed' this.  I'm not certain exactly what
   happend, tho.

4) I currently have 11 separate patches.  Most are not debian-specific.  I'll
   send those soon as well, but acceptance is not required before I upload.

   Here's the directory listing; some may be able to take a guess as to what
   each patch does.

5) I see it's not possible to have both a xen microkernel that supports both
   pae and non-pae.  For individuals who are compiling themselves, that's
   fine; they can select which they want.

   However, for debian, I need to provide versions of xen compiled for both.
   The build system doesn't make that as easy as I like.  However, I'll fix it
   one way or another, even if I brute-force copy the entire directory.

   Also, what other options are available, that would require multiple
   compiles?  Also, I don't see a concise list of what config options I can
   tweak for xen itself; the documentation is a bit poor in that regard(pae
   isn't even mentioned in the documentation; you have to read a separate
   README).

6) This isn't a big problem, and I'll probably just punt on it(by having the
   xen3 debs conflict with the xen2 debs as appropriate), but it's not
   possible to install xen2 and xen3 support code at the same time.  All of
   the c binaries, and python scripts could be made to look in versioned
   locations; however, the python modules are the problem.  It'd be really
   nice if instead of placing themselves in a xen directory, they were in a
   xen2 or xen3 directory.

I won't be doing much of any of the work above today.  I only had 3 hours
sleep last night(had to rush-deploy the xen3 server we were preparing, so we
could migrate a broken real machine to xen).  However, this weekend will see
me upload to debian itself.  I'll also be attempting to make the deb work on
amd64 and ia64, doing test compiles on debian's machines first.

Re: [Xen-devel] Xen 3.0.0 released!

[Ewan Mellor]

On Wed, Dec 07, 2005 at 04:36:11PM -0600, Adam Heath wrote:

> > > 2) vif-route.sh is buggy; it doesn't handle the case where
> > > multiple adapters
> > >    exist.  So, dom0_ip(and the functions it calls) ends up
> > > returning 2 ip
> > >    addresses, which makes ifconfig complain.
> > >
> > >    I've got a fix hot-patched on our deployed server, I'll be
> > > sending that
> > >    soon.
> >
> > Please post it, thanks.
> 
> In vif-common.sh, ip_of should be this:
> ==
> function ip_of()
> {
>   ip addr show "$1" | awk "/^.*inet.*$1\$/{print \$2}" | sed 's,/.*,,'
> }
> ==

Applied, thanks.

Ewan.

RE: [Xen-devel] Xen 3.0.0 released!

[Ian Pratt]

> I've had 3.0-unstable debs prepared for awhile now; hadn't 
> uploaded them, for several reasons:
> 
> 1) xen-unstable required a 2.6.12 base kernel; this no longer 
> exists in
>    debian, so it wasn't possible for me to build-depend on 
> it, so I could
>    create a standard unified-diff.

Do you have 2.6.14? You could build based on linux-2.6-xen.hg, but this
hasn't has s much testing (we know of at least on issue on x86_64)
 
> 2) vif-route.sh is buggy; it doesn't handle the case where 
> multiple adapters
>    exist.  So, dom0_ip(and the functions it calls) ends up 
> returning 2 ip
>    addresses, which makes ifconfig complain.
> 
>    I've got a fix hot-patched on our deployed server, I'll be 
> sending that
>    soon.

Please post it, thanks.
 
> 3) It appeared that when I tried the final 3.0.0 release, 
> that xenstored was
>    very sick.  I had a previous(20051116 snapshot) installed, 
> and the tdb file
>    was corrupted(sorta).  xenstore-list / shows to local 
> paths, but no vm nor
>    tools.  Removing the file 'fixed' this.  I'm not certain 
> exactly what
>    happend, tho.

It's possible that the install scripts should delete the database when
performing an upgrade. I don't think there have been any format changes
in a while, though.

> 5) I see it's not possible to have both a xen microkernel 
> that supports both
>    pae and non-pae.  For individuals who are compiling 
> themselves, that's
>    fine; they can select which they want.

Yep, no real way round that. Most users are already trained to this
thanks to it being a compile time option in Linux.
 
>    Also, what other options are available, that would require multiple
>    compiles?  Also, I don't see a concise list of what config 
> options I can
>    tweak for xen itself; the documentation is a bit poor in 
> that regard(pae
>    isn't even mentioned in the documentation; you have to 
> read a separate
>    README).

Having a debug/verbose build of xen is sometimes handy for debugging
things, but since debug builds are deliberatly combatitive its probably
not a good idea for production. 

Ian