[OmniOS-discuss] oracle vm server in kvm enabled zone

Jim Klimov jimklimov at cos.ru
Thu Aug 11 20:46:38 UTC 2016


11 августа 2016 г. 12:36:35 CEST, pieter van puymbroeck <pietervanpuymbroeck at hotmail.com> пишет:
>Hi Jim,
>
>
>Basically it's a personal experiment. I'm an oracle dba and I test
>everything in my home-lab before talking to customers.  I'm very used
>to ovm. But I can't use that because I want to stick with one physical
>box, so the kvm enabled zones are exactly what I need.
>
>
>You're 100% right that everything I want to do it possible directly in
>kvm as well except one thing. A self service portal on which I can
>prepare my environments and that they get cleaned up automatically
>after the tests, that's the main drive to use ovm.
>
>
>I do like the solaris (well omnios) zones that much that I use them for
>the rest of my home network (web server, dns, mysql,... the usual
>stuff) but at this point I did not search good enough I think, but I'm
>looking for something that I can click in a webpage "I want such a vm,
>that os, that storage, that oracle version and a test db yes or no".
>
>You're right I can start scripting it by building a pxe server,
>preparing the images and build a bunch of ansible / puppet / bash
>scripts which kick off when needed, but it's a lot of work and I was
>thinking, I can build it easily with cloud control communicating to
>ovm, so why shouldn't it work.
>
>And to a certain height,it does work. The ovm servers are running
>wonderfully stable, except the kvm's don't allow nesting and that's
>what I need [��]
>
>
>If you have alternatives for such kind of selfservice portal (personal
>use, so must be quick and easy to configure!!!) I love to hear it, then
>the problem is solved as well.
>
>Best regards,
>Pieter
>________________________________
>From: Jim Klimov <jimklimov at cos.ru>
>Sent: Thursday, August 11, 2016 8:57 AM
>To: pieter van puymbroeck; Jim Klimov
>Cc: Michael Rasmussen; omnios-discuss at lists.omniti.com
>Subject: Re: [OmniOS-discuss] oracle vm server in kvm enabled zone
>
>10 августа 2016 г. 19:51:38 CEST, pieter van puymbroeck
><pietervanpuymbroeck at hotmail.com> пишет:
>>Clear.
>>So I don't have to expect it soon.
>>Pity, because I love this Solaris fork and besides hobying (for work),
>>a use case might be to make abstraction of the underlying architecture
>>(Intel/amd/...).
>>But ok, I didn't make a mistake it's just not possible using omnios,
>>right?
>>
>>
>>
>>
>>
>>> Op 10 aug. 2016 om 19:12 heeft Jim Klimov <jimklimov at cos.ru> het
>>volgende geschreven:
>>>
>>> 10 августа 2016 г. 16:52:56 CEST, pieter van puymbroeck
>><pietervanpuymbroeck at hotmail.com> пишет:
>>>> Isn't it supported in omnios?
>>>>
>>>> It is supported in different linux flavours, but I don't like to go
>>on
>>>> that way.
>>>>
>>>>
>>>> I read on the omniti website that kvm is one of the key-features,
>so
>>>> nesting was something which I'd expected to be there. Don't get me
>>>> wrong, the not nested kvm is working perfectly (and fast!).
>>>>
>>>>
>>>> Is there an estimated timeline in which it should be supported or a
>>>> place to launch enhancement requests?
>>>>
>>>>
>>>> ________________________________
>>>> From: Michael Rasmussen <mir at miras.org>
>>>> Sent: Wednesday, August 10, 2016 2:48 PM
>>>> To: pieter van puymbroeck; omnios-discuss at lists.omniti.com
>>>> Subject: Re: [OmniOS-discuss] oracle vm server in kvm enabled zone
>>>>
>>>> Seem to remember nested kvm is not supported.
>>>>
>>>> On August 10, 2016 3:53:11 PM GMT+02:00, pieter van puymbroeck
>>>> <pietervanpuymbroeck at hotmail.com> wrote:
>>>>
>>>> Hello,
>>>>
>>>> Currently I'm building a new homelab based on omnios
>>(omnios-r151018)
>>>>
>>>> There a kvm enabled zone is created and inside the zone I have 3
>>>> kvm-guests running. These guests are running oracle vm (xen based).
>>All
>>>> is running fine except if I want to start a linux-testvm inside
>>oracle
>>>> vm, the system complains about:
>>>>
>>>>
>>>> Error: HVM guest support is unavailable: is VT/AMD-V supported by
>>your
>>>> CPU and enabled in your BIOS?
>>>>
>>>>
>>>> Inside the kvm enabled zone, kvm is enabled:
>>>>
>>>> root at ovm1:/root# ls -l /dev/kvm
>>>>
>>>> crw------- 1 root sys 125, 0 Aug 10 12:48 /dev/kvm
>>>>
>>>> root at ovm1:/root# modinfo |grep -i kvm
>>>>
>>>> 207 fffffffff840a000  3a030 125   1  kvm (kvm driver v0.1)
>>>>
>>>> root at ovm1:/root#
>>>>
>>>>
>>>> For sure I checked the processor itself from the global zone if it
>>is
>>>> supported:
>>>>
>>>> # isainfo -v
>>>>
>>>> 64-bit amd64 applications
>>>>
>>>> vmx pclmulqdq aes sse4.2 sse4.1 ssse3 popcnt tscp cx16 sse3 sse2
>>>>
>>>> sse fxsr mmx cmov amd_sysc cx8 tsc fpu
>>>>
>>>> 32-bit i386 applications
>>>>
>>>> vmx pclmulqdq aes sse4.2 sse4.1 ssse3 popcnt tscp ahf cx16 sse3
>>>>
>>>> sse2 sse fxsr mmx cmov sep cx8 tic fpu
>>>>
>>>> #
>>>>
>>>>
>>>>
>>>> and to me it looks I have the required flags available.
>>>>
>>>> inside the lvm-guest the flags are indeed not given to the guest:
>>>>
>>>>
>>>> [root at ovm2 ~]# uname -s -r -i -p
>>>>
>>>> Linux 2.6.39-400.279.1.el5uek x86_64 x86_64
>>>>
>>>> [root at ovm2 ~]# cat /etc/oracle-release
>>>>
>>>> Oracle VM Server release 5.7
>>>>
>>>> [root at ovm2 ~]# grep flags /proc/cpuinfo
>>>>
>>>> flags : fpu de tsc msr pae mce cx8 apic sep mca cmov pat clflush
>mmx
>>>> fxsr sse sse2 ss syscall nx lm constant_tsc up nopl pni ssse3 cx16
>>>> sse4_1 sse4_2 popcnt aes hypervisor lahf_lm
>>>>
>>>> [root at ovm2 ~]#
>>>>
>>>>
>>>> it does not matter if I start the kvm guest using
>>>>
>>>> -cpu host \
>>>>
>>>> or using
>>>>
>>>> -cpu qemu64,+vmx,+aes,+sse4.2,+sse4.1,+ssse3 \  # --> this one is
>>>> faster
>>>>
>>>>
>>>> despite I have an intel cpu I use the -enable-nested flag as well,
>>but
>>>> it doesn't seem to make a difference.
>>>>
>>>>
>>>> How can I fix this?
>>>>
>>>>
>>>> Thanks and best regards,
>>>>
>>>> Pieter
>>>>
>>>> ________________________________
>>>>
>>>> OmniOS-discuss mailing list
>>>> OmniOS-discuss at lists.omniti.com
>>>> http://lists.omniti.com/mailman/listinfo/omnios-discuss
>OmniOS-discuss Info Page - lists.omniti.com Mailing
>Lists<http://lists.omniti.com/mailman/listinfo/omnios-discuss>
>lists.omniti.com
>OmniOS is a minimalist, self-hosting, Illumos-based distribution aimed
>at server deployments. Release and update announcements are also made
>to this list.
>
>
>
>>>> OmniOS-discuss
>>>> Info Page - lists.omniti.com Mailing
>>>> Lists<http://lists.omniti.com/mailman/listinfo/omnios-discuss>
>>>> lists.omniti.com
>>>> OmniOS
>>>> is a minimalist, self-hosting, Illumos-based distribution aimed at
>>>> server deployments. Release and update announcements are also made
>>to
>>>> this list.
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Sent from my Android phone with K-9 Mail. Please excuse my brevity.
>>>>
>>>>
>>>>
>>------------------------------------------------------------------------
>>>>
>>>> _______________________________________________
>>>> OmniOS-discuss mailing list
>>>> OmniOS-discuss at lists.omniti.com
>>>> http://lists.omniti.com/mailman/listinfo/omnios-discuss
>>>
>>> AFAIK this is a common limitation with x86 HW hypervisors that only
>>the barebone system may grab and lock the CPU extensions that are the
>>HW hypervisor implementation. Guest OSes may only do unaccelerated
>>software emulation for their sub-guests.
>>>
>>> For example, to use VirtualBox on Windows (Servers) you must ensure
>>that HyperV is disabled and its drivers are not loaded at all.
>>>
>>> Jim
>>> --
>>> Typos courtesy of K-9 Mail on my Samsung Android
>
>Hi Pieter,
>
>Is there a particular reason for you to stack hypervisors? Why run
>guests under OVM under KVM, and not guests in KVM directly? (Or if
>compatibility of this KVM port is insufficient for your usecase,
>perhaps try VirtualBox instead)?
>
>As for "others" - this is the first time I hear of having accelerated
>second-layer hypervisors, so the rest is my educated guesswork and pure
>speculation. Granted, I do not have much experience with KVM nor Linux
>hypervisors lately... so my guess would be that certain code
>intentionally plays well together (e.g. top-level KVM providing or
>wrapping access to CPU VT-* extensions to guest KVMs), but this is not
>likely a universal solution that lets everyone play with anyone (KVM,
>HyperV, VirtualBox, VMWare, ...), that Linux or everyone else has and
>only Solaris/illumos lacks. Or maybe a lot happened in upstream since
>the port was done ;)
>
>Jim
>--
>Typos courtesy of K-9 Mail on my Samsung Android

You can run Oracle DB for Solaris in zones ;) For the installer you can set up X11 (+vnc on server, or direct for workstation) and use `xhosts` to propagate the session into local zones.

For tests and shows you can pre-install a zone with maybe an empty database, shut it down, and then clone and tear down for practical short-lived purposes. Being zfs clones, this would involve minimal overhead on storage side.

Likewise, you can clone golden datasets with virtualbox vm's and rewrite virtual disk id's (documented) and machine manifests to attach a new/cloned vm to such cloned disk. Likely similar approach is doable with KVM. But this is a bit of scriptable or manual work...

But sorry, no point-and-click magic that I'd know of...

Jim
--
Typos courtesy of K-9 Mail on my Samsung Android


More information about the OmniOS-discuss mailing list