[OmniOS-discuss] Reversed All-In-One

[Jim Klimov]

25 сентября 2016 г. 16:36:26 CEST, Paul Jochum <paul.jochum at nokia.com> пишет:
>Hi Michael:
>
>     As to your 2nd qustion, about having good info on how to create a 
>custom lx image, I have used the following for smartOS: 
>https://github.com/joyent/centos-lx-brand-image-builder
>     (I am not sure how closely OmniOS tracks the smartOS lx branded 
>zones, but wanted to pass it along in case it helps).  There are more 
>links (in case you don't want to use centos as the base at:
>https://wiki.smartos.org/display/DOC/LX+Branded+Zones#LXBrandedZones-Creatingyourownzonedataset
>
>Paul
>
>
>On 09/24/2016 02:00 PM, Michael Talbott wrote:
>> So I just had this crazy idea while thinking about lx branded zones..
>I'm loving lx zones. What a great piece of computing voodoo. I'd love
>to say there's no more need for things like ESX in an all in one setup,
>but, there's still windows and some random others that has their niche
>needs on a box.
>>
>> Anyway, I know there's plenty of folks out there that use ESXi as the
>root system then virtualize OmniOS and pass through storage controller
>cards to it to become a storage VM for their other VMs. I started
>thinking "what-if" we reversed that process to get the best of ALL
>worlds. That is, run OmniOS as root system and control storage natively
>and run all linux needs in zones. Then if, and a big if, we could run
>ESXi in an lx zone.. And IF that is even possible, another big if.. IF
>we can present a zone with a ZFS volume as a raw block device, ESXi
>could then use it's native VMFS format on said block device. If that
>were so, that would eliminate all traditional latencies and bottlenecks
>caused by the network stack be it iSCSI or NFS and provide all the
>benefits of ZFS such as checksums and compression inline to ESX..
>Obviously it couldn't vmotion in a cluster since the storage would be
>local to that box, but, the thought is currently for an all in one
>setup and not a clust!
> er!
>>    ;)
>>
>> So I guess the first question is, does anyone think that ESXi's
>kernel is anywhere close enough to a linux kernel for this to even be
>feasible to run in an LX zone or am I totally just daydreaming here? I
>have my suspicions, but, that's all they are at this point.
>>
>> Second, and regardless of the above, does anyone have any good info
>on how to create a custom lx image? Surely it's not as simple as
>installing it on some box and then use tar and/or dd?
>>
>>
>> Michael
>> _______________________________________________
>> OmniOS-discuss mailing list
>> OmniOS-discuss at lists.omniti.com
>> http://lists.omniti.com/mailman/listinfo/omnios-discuss
>
>
>_______________________________________________
>OmniOS-discuss mailing list
>OmniOS-discuss at lists.omniti.com
>http://lists.omniti.com/mailman/listinfo/omnios-discuss

As far as I know, ESX used a custom kernel as hypervosor (closer to BSD than Linux, perhaps because the latter would have to get opensourced). The linuxish environment seen in older (3.x) versions and hidden but uncoverable in newer ones is a controlling VM (IIRC, may be wrong here) allowed to grab the console and control the HW resources to present them back to hypervisor, and indeed based on Linux (quite compatible to redhat back then). To the point where I could compile kernel modules like md-raid and drivers to use setups other than certified officially.

I guess the entanglement there is too intimate to port elsewhere, and why would you - given that you can run KVMs or VirtualBoxes natively in a global or local zone? ;)

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