On 11/08/2009, at 7:13 PM, Anon Y Mous wrote:
Please, you can't discontinue SXCE because OpenSolaris Indiana
2009.06 isn't ready yet as a "headless" minimal server OS that I can
install and manage entirely through a DB9 to RJ45 CISCO console
cable (think about the "Netra T1" and all of the other servers Sun
makes that don't have graphics cards in them).
I know this for a fact because I have been using every major Indiana
release from 2008.05 snv_86 to 2009.06 snv_111 as my desktop OS on a
daily basis (I will also have Indiana running full time on the
Toshiba Tecra laptop after tomorrow as well) and IMO Indiana is
missing many of the essential features that sysadmins need for
servers like flash archive / flarcreate for backing up and restoring
servers, the ability to assign a static IP address to the server
during the installation process (and not be forced to use NWAM /
DHCP), and the ability to have a tty console cable text installer do
a minimal server installation (with only SSH running in the global
zone after the install).
You should have a look at -
http://www.opensolaris.org/os/project/caiman/TextInstallerProject/
It will solve many of the issues you're seeing with installation. I
know the Caiman folks are also working on replication and recovery
project to help fill the FLAR gap - unfortunately I can't easily find
details of that project, but hopefully one of the team will pitch in.
Live Upgrade is also probably much better for mission critical
servers than "pkg image-update" is in Indiana, because if your root
file system gets corrupted and you have live upgrade set up
properly, then there is a chance that you can still boot up into the
alternate slice and still keep your sysadmin job!!!! If you're an
OpenSolaris Indiana sysadmin and you happen to lose your "rpool"
root zpool like some of these poor guys did (see links below):
http://opensolaris.org/jive/thread.jspa?threadID=108722&tstart=45
https://www.opensolaris.org/jive/thread.jspa?threadID=108213&tstart=30
https://www.opensolaris.org/jive/thread.jspa?threadID=90432
then you are S.O.L. and you will get fired and all your Sun kit will
probably get junked and replaced with IBM servers running Red Hat
Enterprise Linux backed up by Netapp filers. Pkg image-update is
nice and all, but if I had to bet my job, my career, my livelihood
and the future of the Solaris OS on a mission critical server that
absolutely CAN NOT FAIL, I would bet it on Live Upgrade.
If you really cared about corruptions on disk, you'd be using some
sort of ZFS mirroring/RAID-Z.
To use OpenSolaris Indiana as a minimal server OS that can be
managed from a a DB9 to RJ45 CISCO console cable, I would have to do
what Milax or Nexenta does and build my own minimal OpenSolaris
server distro from scratch from the source code, and this is too
much work and too much liability for the average sysadmin to do.
Plus doing this gives me a configuration that is not supported by
Sun, so Sun makes no money from me using OpenSolaris because I can't
buy "server support" for it from Sun and my boss and co-workers are
angry because we're taking on the burden of building our own
unsupported operating system from scratch.
I'd be curious as to what you'd expect "server support" to be - are
you talking about the length of subscription, price, or content?
The end result of this situation of killing off SXCE would be all
Solaris Express-using server customers giving up the Solaris OS and
changing to another OS and buying support from Canonical or Red Hat
or some other vendor that supports minimal server configurations
that don't have X windows installed on them, which would be the end
of Solaris, which would be very sad because Solaris has a lot of
good features that don't exist in other operating systems.
That's a weak trail of thought at best. I appreciate there are missing
features that we know our enterprise customers are likely to want and
need, and we're not quite there yet. However, they can continue to use
Solaris 10 as a fully supported enterprise option as a backstop. The
gaps will be closed real soon. The interactive text install will have
a different default package profile to be installed by default, and
won't include a graphical desktop environment on current thinking.
You can't expect to make any revenue for Sun having your main
product be something like Indiana that only works well on desktops
and not on headless servers. OpenSolaris Indiana 2009.06 is not
going to get a majority desktop share over Microsoft Windows and Mac
OS X any time soon, so by killing of SXCE (the only "server
friendly" version of the OS that has the cool features like crossbow
and sparse root zones in it), you're killing of Solaris :-(
I think you're mistaken about where things are going over the next 6
months or so. I'd expect 2010.02 to move pretty heavily towards
providing functionality for server environments - on disk IPS support,
replication and recovery, hardened Automated Install, Solaris 10
branded zones, etc. OpenSolaris OS is 'server friendly' right now -
you just have to do a little additional work to get there, and some of
the barriers to entry are a little higher than we would have liked.
Wouldn't it just be easier to keep SXCE going for another year and
pick a few random "stable" builds that we can buy support for like
we could with SXDE? I miss SXDE. SXDE was awesome :-(
Help all of us in the OpenSolaris community to help Sun by giving us
a stable SXCE / SXDE build that we can buy support for. This will
generate revenue for Sun and the revenue can get re-invested back
into the OpenSolaris project R&D to make Indiana an even better OS.
By continuing to support SXCE/SXDE, you end up distracting from the
overall work we need to do to OpenSolaris. We currently support
OpenSolaris OS - you can buy a subscription whether you are running it
on a workstation or headless server - and for many, this is more than
enough (especially so given the package content is so similar between
the two trains).
Glynn
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org