Kent Watsen wrote:
Can I consolidate all my servers to one machine - here are my current
machines:
1. OpenBSD (used for external services: dns, http, smtp)
2. OpenBSD (used for internal services: dns, http, smtp, imap, ldap,
smb, nfs, svn, bugzilla)
3. OpenSBD (used for upgrading either of the above servers without
any downtime)
4. OpenBSD (used as a build/test machine)
5. FreeBSD (used as a build/test machine)
6. NetBSD (used as a build/test machine)
7. RedHat (used as a build/test machine)
8. CentOS (used as a build/test machine)
9. SuSE (used as a build/test machine)
10. Solaris (used as a build/test machine)
11. Windows (used as a build/test machine)
12. MacOS X (used as a build/test machine)
Pretty interesting, as this is more or less the exact same situation
we're having at my job - building for multiple platforms.
We're in the early planning stages on how to solve that scenario, and
I've been sysadm for only 1½ month, so a Bit of time's needed ;)
You say build platforms, but don't mention testing.
Our software is OpenGL/3D dependant.
We have two scenarios: Building and testing.
It's mostly for XP, RH, Suse and OSX, + other *nix's to some extend.
We may choose to crosscompile as much as possible from a couple of
boxes, and use a virtualized setup for testing. At the moment, I can't
see Xen used for testing, due to inappropriate graphics in domU's, so we
may have to use vmware, though it's not the fastest.
A few comments:
ad 12: I know OSX can install on a emulated X86, but I don't have
personal experiences. I would doubt it's usefullness. At least we're not
going to drop our G5 for that :)
ad 11: I'd wait for AMD in june, but still, graphics in Xen?
ad 10: Don't know the status of OpenSolaris on Xen. One domU instance
should work, apart from that, dunno.
BSD's: Same as ad 10. AFAIK, all should install in domU, but how well
they run, dunno. Would like a refresh in this. Anyone?
Your listing seems to suggest you're basically BSD based.
Planning to use one of the BSD's for dom0?
If you consolidate /all/ your services on one box, you'll create a
single point of failure issue for yourself.
I would at least use one virtualized box for infrastructure and another
for the building.
Having two identical all-in-on boxes with redundancy/failover might be
another solution. Might even be used for parallelized builds...
And maybe keep a dedicated Mac.
I'm having a related situation at home. Having stopped freelancing and
gotten a good job, I just wan't a single box here.
Been looking a Shuttle's lates announcement, an SFF box for AM2, taking
4GB DDR2, using nVidia 51-series chipset.
Might be an idea for us to keep in touch on this.
Notes:
* The OpenBSD-based servers are RAID-ed
* There are actually more machines as I run multiple releases of
each build/test OS...
If it can't be done on one machine, than would either of these 2-machine
solutions work:
1. Partition machines by server vs. build/test
* one machine has: 1-3 (all para-virtualized)
* other machine has: 4-12 (5/9 para-virtualized)
2. Partition machines by para- vs. full-virtualization
* one machine has: 1-4 and 10-12 (all full-virtualized) [would
GSX be better?]
* other machine has: 5-9 (all para-virtualized)
What would you do?
Thanks!
Kent
------------------------------------------------------------------------
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
--
Kind regards,
Mogens Valentin
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
Home