XenServer 8.0 - Major update due Q1 2019
-
@jcpt928 said in XenServer 8.0 - Major update due Q1 2019:
@cg It has long ago become apparent to us at work, as long-time paying customers of Citrix, that they have little interest and long-term motivation towards XenServer\Citrix Hypervisor, much less a number of their other products, despite the way they brag about them in relation to other industry solutions. They seem more focused on squeezing as much money out of existing customers than actually providing, and then walking, any seemingly beneficial road map for the future that consists of real growth, new ideas and concepts, and expanding beyond "what they already know" to provide new and existing customers with improved solutions.
I'd like to quote this post because this was also one of the fork reason. I mean, even from a business perspective for Vates, it made sense to deliver innovation and a good service, which both aren't provided anymore for server virtualization at Citrix. Obviously, we are not that big, but unlike them, we are really open from the start (cultural difference) and also not being pushed by big pension funds (remember that squeezing the money starts at this level first). I still have in mind those big funds who pushed out the historical CEO, Mark Templeton. Which was the guy with the vision
(note: he's now CEO of… Digital Ocean!) -
@cheese said in XenServer 8.0 - Major update due Q1 2019:
Should XCP-ng aim to be upstream? Like Debian/Fedora, XCP-ng maintains an unstable/cutting-edge branch that directly follow each XCP package?
I'm not sure to understand what you mean by that, can you explain further or rephrase?
-
@olivierlambert Sorry. Currently XenServer is the upstream of XCP-ng. And XCP-ng does not directly choose XCP package versions. XCP-ng will not make a new major release until XS released new version. How this can be improved? Can XCP-ng get rid of the XenServer stage? XenServer itself is composed from XCP packages or is a distribution of XCP. Can XCP-ng directly be based on XCP packages instead of XenServer? Or for comparing, should XCP-ng be the 'Fedora' of XCP instead of the 'CentOS' for XS?
Of course, being 'Fedora' of XCP requires a huge work. And QA is a big problem.
-
Well, that's not true @cheese
The truth is far more complicated:
- XCP-ng already added some new packages inside (eg zstd, modified XAPI to support it, a more recent kernel etc.)
- Some bug are fixed first in XCP-ng (and later in XS by Citrix)
- "XCP" doesn't exist anymore (since 2013), so we can't use it as a base
- Some packages aren't shipped until XS is out and no dev branch is available (thinking about SMAPIv3): it's legal but really anti-Open Source spirit (you can't contribute)
- Some packages are heavily modified by Citrix, eg Xen (400 patches…) so you can't "just" import any vanilla Xen into XCP-ng. This require months of deep Xen knowledge.
But yes, the goal is to be 100% independent in the end, but that require a reasonable team (already hired 3 persons, objective is to double this number in the next months) and as you can imagine, this cost a lot of money. So break even is needed to be sustainable. That's why being only "Fedora like" (eg cutting edge without support) isn't doable. We must also provide something tested and stable through time, otherwise no company will adopt it, and then we couldn't survive (Patreon or Kickstarter is OK to bootstrap a project, but just add 6 to 10 people full time and you'll see you'll need something close to one million USD and more)
So as you see, XCP-ng is not just a toy done in a garage, this is bigger than that
Our ambition is to become the first turnkey and Open Source virtualization platform. Some parts will take time but yes, this is the objective. -
@olivierlambert said in XenServer 8.0 - Major update due Q1 2019:
Our ambition is to become the first turnkey and Open Source virtualization platform.
And the community involved here is progressing to achieve this collectively. -
Yep, "our" was inclusive
-
This is standard Citrix.. I don't see XS8.0 showing up until Synergy next month. They've already announced the next Xendesktop/XenApp CR and LTSR release this month a so they will need a new product release for the show. I had hoped that they would have released a Tech preview before the full release just to see if it'll work on existing older kit.
-
It's not a marketing delay. They have some issues, the target was end of Q1 for real.
-
@olivierlambert Oh! I appreciate your plan.
And I am sorry again. "XCP" in my previous messages actually means XAPI project. Sorry.
-
@olivierlambert which is why you, sometimes, better just release and deliver unfinished stuff with an update - unless it's some very important core funcionality.
Also I don't expect 8.0 to be the next LTSR - IIRC they want to bring that one later in the year, which would mean: Most people won't care, if they deliver something (a bit) later.I hope XCP-ng will become independend, but for now I see a lot of work and money...
I simply can't bring any licensed models to some smaller customers, as the fees are too high, relative to their pretty small servers. So an option to pay a smallter amount, idially a one time fee... maybe you bring something on the road (same applies to XOA). -
Clearly, in a niche market like this, very cheap one time fees aren't a viable solution to survive (because the market is too small). Cheap but recurrent is already more doable (hence XOA Starter for example, or XCP-ng Standard).
Also, if you really can't spend a dime on it, use it without support for free (and be the support if needed). And ideally, contribute
To give you some hints with basic maths, when you need approx 100k€ per month for a decent dev team, how many recurrent XOA Starter do you need to pay for it? vs how many Premium? You can see one of those 2 is achievable, the other is really harder because of the market size. And it's even worse with a one time fee: you need the same amount of new customer every year, not to earn money, just to cover your expenses. Take this info in perspective, this will give you a better view about the cost to develop software
-
@olivierlambert I know what you mean, but other companies have that offers and survive, too. Of course the lower priced offers are more restricted, but that's fine for small companies. I do contribute here and there to some things, but my time isn't endless, nor is my knowledge.
I'm admin, not a coder. -
- Other companies aren't giving flat pricing for their backup solution
- XOA isn't just a backup solution
- Other solutions aren't Open Source
So as you can see, it's very different
About contributing: already doing some specs (how to access a tape drive?) would help to improve XO.
-
- True, but when your backup is a few 100 GB... you don't care.
- They don't need the rest, as there's no other HW anyways.
- This argument won't make much companys pay thousands of € extra, over the years.
I absolutely have no clue how tape drives work, software wise.
Telling you hardwarespecs of tapes, drives and versions and compatibility most likely won't help you.
I can use it, and maybe, I'd could do the efford to test stuff on older hardware and give feedback - but for now, that's it.Edit: I saw that HPE supports you, so maybe you have a chance to get informations form them, as they sell tape drives (HPE Ultrium/StoreEver... LTO) and should be able to support you with informations. They helped me once with configuring XS for use with HPE MSA.
-
I will add that I use XCP-ng at home; but, also in an enterprise virtual environment (without professional support right now, as we don't need it for Xen). We also use the Enterprise XOA, and have even had great success with Backup-ng for specific VMs we can't backup with other solutions. My qualm is that I could absolutely use some of those Enterprise features at home; but, alas, the licensing is WAY beyond my personal budget. The subscription model is the only way to go in this industry, or the very similar perpetual license with a yearly support subscription (neither all that different from each other from a cost perspective) - you can't maintain a team and ongoing development with "one time fees" (look how this has actually harmed Plex due to way too many lifetime passes being sold).
That said, the reliability and stability of the XCP-ng\XOA solutions across the board is at least equivalent to native Xenserver (which isn't necessarily an insult or a compliment - any of us who have used XenServer long enough know that it can be incredibly easy and downright simple to manage, or it can turn into an absolute disaster real quick). We also run native XenServer, and some VMware and Hyper-V in our environment.
Do I have ideas of where I'd like to see XCP-ng go? Absolutely. I don't think it's time for me to be pushing those while the team fleshes out the product per their initial goals and desires, however. Do I have extremely diverse experience that could be invaluable to the team in developing new features and capabilities? I'm sure; but, I'm not a developer, I'm an administrator\engineer. I understand coding, and can often talk the talk; but, it's not my cup of tea, for sure - I'm not sure I'd know where to start to assist those endeavors beyond asking actually questions.
I do hope that the XCP team's recent progress towards becoming an independent product is a success - sooner rather than later. They already have some big companies betting on them. I could go on for quite a while on subjects like this - if XCP-ng has an open position that pays roughly $100K/year for someone in the top 5% of skill and talent in the global IT\IS field, let me know.
-
@jcpt928 if you can bring support contract for at least 170 XCP-ng "standard" hosts per year, then your cover our expenses to pay you
-
-
I have installed the Citrix Hyperivsor 8 and recompiled your xcp-featured Package and modified it a little bit for the new Features like UEFI Secureboot. It seems to be working so far. I just use this for my Homelab.
-
Feel free to share your modifications
-
I have removed the Feature Restricton of Corosync, that it is also available, because it is availbale in the Citrix Hypervisor 8, but I think you aren't allow to distribute this Feature, as it isn't Open Source. I wanted to test the GFS2 Feature.
I have added Corosync in additional_feature and in keys_of_additional_features the same I also did for the UEFI Secureboot.
I added it in additional_feature with the Name "GuefiSecureBoot" and in "keys_of_additional_features" with "GuefiSecureBoot, (Negative, "restrict_guefi-secureboot");"I hope it is clear what I did.
I forgot to save the patch File otherwise I could send it to you. The Compilation of the Package was done with your xcp-ng-build-env.
