Running away screaming from ESXi (Free), best alternatives?

Innocence

2[H]4U
Joined
Mar 9, 2001
Messages
2,604
Hi All,

I've been working with virtualizing quite a bit over the last couple years, and adopted vmWare ESXi pretty quick. But with the release of ESXi 5.5, it's painting a pretty grim picture of the direction that vmWare is heading, especially with their free releases.
The main use I'm getting out of my ESXi servers is to p2v old, seldom used servers that we have to keep around for "legacy support" or as a catch for clients who haven't been moved to a newer system (roadwarriors who are never in the office, clients who haven't been migrated, etc...). Cleared up about 80U of rack space by killing off towers and massive old 6U servers.

I haven't virtualized any essential infrastructure, and likely won't do so anytime soon for a sack full of reasons. But as it stands, I have about 25 VMs running on 4 host servers, all ESXi (free) 5.1 (U1).

So almost certainly moving away from vmWare soon, but I do need a stable platform to continue working with, and to support testing environments. Anybody have good suggestions or horror stories?

Going to try :
Xen - With XenServer going "Open" it's certainly on the short list. Lack of OpenBSD support (and documented issues with networking) is a big concern though since I run a few dozen OpenBSD machines.

HyperV - seems an obvious choice for straight Windows Servers, but with Microsoft I always feel like they're waiting for everyone to get in the pool -- then they drop in a money-eating shark.
 
Well, aside from them moving over to a "VaaS" model, which let's face it, EVERYBODY is doing including Microsoft... what, exactly, has you painting a grim picture of VMWare's desires? I currently see nothing other than their aforementioned services model (again, which EVERYBODY is going to try to be pushing you towards here shortly) that 5.5 is doing that is different from previous versions.

Their only cock up imho is that they're trying to push everything to a web based management system, but neglected to tell end users of the free product that 5.5 does not have any free option for web based management. Supposedly that will be rectified down the line though.

Still, I'm curious anyway.
 
My guess is that what the OP is talking about is the lack for whitebox hardware support with the removal of Realtek NIC drivers from 5.5.

Anybody have good suggestions or horror stories?

You fill find an equal amount of those for all virtualization products out there.

If you are running a commercial operation then the cost of virtualization software licensing is the cost of doing business. If the business isn't viable unless you run some open source license fee free solution then there's a problem with the business, not with the cost of the technology.

If you are using the free version of ESXi then it really doesn't matter what other product you switch to because the free versions of the competitors are all essentially equal. On the commercial side nothing comes close to VMware vSphere in a mixed guest OS environment.

On a more personal note; I really take issue with folks who want free stuff. I don't work for free, I doubt many people do, good software and services provide value and thus it's a non-issue if those require payment.
 
On a more personal note; I really take issue with folks who want free stuff. I don't work for free, I doubt many people do, good software and services provide value and thus it's a non-issue if those require payment.
Open is most important - not free (as in beer). We've made over $100k in contributions to multiple open-source projects over the past decade.
There's also a reason vmWare offers a free version - and it isn't because they're charitable. They have willingly broken a lot of features of 5.5 in the free version by having no vCenter Web interface support - you can't even work with a v10 VM at all without vCenter.

I don't have any problem with paying for software and services that are required and add value - I mean, our budget doesn't spend itself. But I don't believe in the same 'spend like it's on fire' mentality that drives some enterprise IT decisions. Things like this happen when you treat money like it's free

To properly implement vmWare (paid), we are currently quoted $15k for licensing alone. That's outside budget, with a recurring $3k annual mandatory support subscription. For a bunch of systems that are not near business critical (more like "Business Convenient") that's a massive bill. If this was supporting our primary e-mail backend, sure, cost of doing business - but that's not the case here.
 
It seems to me the easiest thing is to stay on 5.1U1 and see what is on the horizon. The landscape with look much different in a year. This allows you to make a better assessment of the company's needs rather than shooting from the hip because of a single update from vmWare. That's like saying Windows 8 stinks when coming from Windows 7 because of "insert any reason here". When in fact 8.1 actually made Win 8 better IMO. To each there own.
 
Open is most important - not free (as in beer). We've made over $100k in contributions to multiple open-source projects over the past decade.
There's also a reason vmWare offers a free version - and it isn't because they're charitable. They have willingly broken a lot of features of 5.5 in the free version by having no vCenter Web interface support - you can't even work with a v10 VM at all without vCenter.

I don't have any problem with paying for software and services that are required and add value - I mean, our budget doesn't spend itself. But I don't believe in the same 'spend like it's on fire' mentality that drives some enterprise IT decisions. Things like this happen when you treat money like it's free

To properly implement vmWare (paid), we are currently quoted $15k for licensing alone. That's outside budget, with a recurring $3k annual mandatory support subscription. For a bunch of systems that are not near business critical (more like "Business Convenient") that's a massive bill. If this was supporting our primary e-mail backend, sure, cost of doing business - but that's not the case here.

That's what I figured it boiled down to. Not that I'm saying you're in any way wrong - I fully agree that by not having a web client for 5.5 "free" while plastering all over the documentation (including for the "free" 5.5 Hypervisor) that such-and-such "requires the web client", yet never explicitly stating anywhere in the documentation that there is no web client for the free standalone host hypervisor was a monumentally huge mistake by VMware with respect to its community.

However, I also see the need for companies to protect revenue streams as well. Remember, the free hypervisor was essentially a response to Microsoft and companys' throw down the gauntlet challenge directly at VMware's incumbency. This is what they do. This is how they generate revenue. You have to look at it from their perspective - can they really afford to cannibalize their flagship product? Microsoft and Citrix can, because their primary revenue comes from elsewhere.
 
I'm not convinced that open is necessarily best. Everyone needs to make a living; I try to remain as objective as I can and pick the one that suits me best regardless. VMware have annoyed me slightly with some of their changes (their MAC address changes screwed up one of my former clients for example) and "forcing" the "web" client similarly annoys (sorry I just like the native one better.) Suppose I'll get used to it though. GUIless Windows 2008 R2 Hyper-V was a sod to setup so hopefully the flipping option makes it easier to setup (paid W2K12 of course). Citrix I see are in a hard place, RDP is getting better all the time - makes XA a hard sell.
 
op can look at
XenServer
Xen Cloud Platform
Linux with KVM
Linux with OSS Xen
Proxmox VE
 
Hyper-V is your answer. If you have an AD domain already you can use Hyper-V core and cluster them together. You'll get HA, Live Migration, Live Storage Migration, and replication for free.

Sure, MS will get more cutthroat as they gain market share but by then I think we'll be so tits deep in public cloud offerings that hypervisors could be a commodity product by then.
 
Open is most important - not free (as in beer). We've made over $100k in contributions to multiple open-source projects over the past decade.
There's also a reason vmWare offers a free version - and it isn't because they're charitable. They have willingly broken a lot of features of 5.5 in the free version by having no vCenter Web interface support - you can't even work with a v10 VM at all without vCenter.

I don't have any problem with paying for software and services that are required and add value - I mean, our budget doesn't spend itself. But I don't believe in the same 'spend like it's on fire' mentality that drives some enterprise IT decisions. Things like this happen when you treat money like it's free

To properly implement vmWare (paid), we are currently quoted $15k for licensing alone. That's outside budget, with a recurring $3k annual mandatory support subscription. For a bunch of systems that are not near business critical (more like "Business Convenient") that's a massive bill. If this was supporting our primary e-mail backend, sure, cost of doing business - but that's not the case here.

You know that the basic version come with 3 host support and its a whopping 600 dollars.

There is no mandatory Support subscription you don't have to pay it if you don't want to. You do have this with MS as well its called Software Assurance except with VMWare you also get tech support thats not fucking located in I can't fucking Understand your Fucked up english land.

As for Hyper V it has most of the features that you want as long as your not running anything but windows and linux if you got legacy like Unix of any flavor don't even bother.

As for the rest of the plebs here yes HyperV is nice but in large environments they are a pain in the ass with out SCCM which is expensive as well.

The reason they are moving to Web because that's what everyone asked for and its less MS licensing you need to get.

I've used all three major Hypervisors. My fav is still VMWare, HyperV is 2nd. VMWare because its easy, HyperV is a clusterfuck to setup and get running correctly with in a decent amount of time and unfortunately its still windows.
 
Thanks for the great feedback.

I know about the essentials license for VMware, but that would severely limit growth and/or require a substantial hardware investment.

The legacy support on hyperv is a concern as well because a lot of the VMs are 2008 and earlier.

Staying on 5.1 is an option for now, but come time for a hardware refresh it could be a nightmare to find supported hardware.
 
That's what I figured it boiled down to. Not that I'm saying you're in any way wrong - I fully agree that by not having a web client for 5.5 "free" while plastering all over the documentation (including for the "free" 5.5 Hypervisor) that such-and-such "requires the web client", yet never explicitly stating anywhere in the documentation that there is no web client for the free standalone host hypervisor was a monumentally huge mistake by VMware with respect to its community.

What can't you do in the C# client with the free version? Pretty sure the web-only features are only in paid licenses.
 
If you only have 25 legacy VMs on 4 hosts....I don't see why you wouldn't just consolidate that to 2 or 3 and get like an Essentials license. Unless those are just monster VMs you're way low on consolidation ratio.
 
What can't you do in the C# client with the free version? Pretty sure the web-only features are only in paid licenses.

Mostly, manage any V10 vdk. Yes, most of the rest of the features in question are paid features at this point. But they also explicitly stated that ANY new features going in from this point forward will NOT be backported to the native client and will only be available via the web client.

Still, a lot of users are getting stuck upgrading their images to version 10 vdk's, and then quickly finding out they have no way to managed them. VMware was silent about that little tidbit from the get go.
 
Mostly, manage any V10 vdk. Yes, most of the rest of the features in question are paid features at this point. But they also explicitly stated that ANY new features going in from this point forward will NOT be backported to the native client and will only be available via the web client.

Still, a lot of users are getting stuck upgrading their images to version 10 vdk's, and then quickly finding out they have no way to managed them. VMware was silent about that little tidbit from the get go.

So don't upgrade to v10. I really, really doubt you'll see anything distributed as v10 by default either. That's it. That's the ONLY thing.

And you'll have a way to manage free hosts from a web client when you need it. Right now you don't need it.
 
25 VMs on 4 hosts...hmmm

We have 500 VMs on 7 hosts...about 130 of which are active.

Our spring consolidation will drop that to 4 hosts to cover 250 active VMs


My plan B upgrade (if we end up finding out we sized it too small) adds another host and splits out our biggest box with independent I/O, supports 400 active VMs adds 7TB of high performance I/O for another $16k.

Imo...You should be looking at Hyper-V. The legacy modes can run pretty much any Linux or windows.
 
If i recall you can also manage HW10 VMs on free ESXi 5.5 via VMware Workstation 10 it has the ability to manage hosts directly and if i recall.
 
So don't upgrade to v10. I really, really doubt you'll see anything distributed as v10 by default either. That's it. That's the ONLY thing.

And you'll have a way to manage free hosts from a web client when you need it. Right now you don't need it.

I completely agree. But the point is it's a pretty crappy precedent to set saying "... all new features will require the Web Client ..." but not inform single host users that the Web Client isn't available, let alone advise them of any caveats as such.
 
I'll tell you guys what when i get a chance ill look on the internal end see if the PM knows about this that way if there isnt something in play we can get the talks going...

I am willing to bet it is known and ill see if there are any workarounds existing to this point and let you know!
 
I completely agree. But the point is it's a pretty crappy precedent to set saying "... all new features will require the Web Client ..." but not inform single host users that the Web Client isn't available, let alone advise them of any caveats as such.

What is the problem? There is no new feature with the free version that requires the web client? Are you complaining because of what might happen in the next version? If so...don't. It's being handled. I can't tell you all the details but stop stressing it.

You're WAY over complicating this. There is absolutely nothing that needs to be configured on the free version that can't with the C# client.
 
Hyper-V 2012. Damn good stuff IMO

I went from vmware to xenserver to hyper-v and really love hyper-v.

I run mostly windows VMs and a couple linux VMs.
 
What is the problem? There is no new feature with the free version that requires the web client? Are you complaining because of what might happen in the next version? If so...don't. It's being handled. I can't tell you all the details but stop stressing it.

You're WAY over complicating this. There is absolutely nothing that needs to be configured on the free version that can't with the C# client.

Junkie they are talking about how Free ESXi builds VMs at HW10 and with the C# client doesnt allow you to edit the settings of a HW10 VM. This Requires the Web Client (vCenter only) sadly or Workstation 10..

Basically the symptoms of
https://communities.vmware.com/message/2292352
 
You don't neven need to do that. You can't build a HW v10 VM with the C# client.

That is true, but iirc you CAN upgrade existing VMs to v10 with the native client (I may be wrong, but that seems to be what several community threads have implied at least), which makes absolutely no sense to be able to do if you can't then manage it with the same client. This is the problem where users are running into, and the point where a lot of users are getting lost as they then go looking for the web client they are told they need only to find out that it's not available for stand alone hosts. And they only find out once they find one of the numerous threads on the community because it's so poorly documented and explained by VMware.

Like I said, I agree with you on principle. This should be a non-issue for most people, and anyone that has taken the time to do a little bit of research first knows that it's a non-issue. As I've said a few times already the problem is readily obvious that in delivering 5.5, VMware did a piss-poor job INFORMING the community of a lot of their ambitions and intentions going forward. It's also leading to people like the OP saying things like he said, given it leaves some question as to what VMware's next moves are exactly going to be with this product.
 
Our organization is 100% VMWare at the moment, but we are looking to move to Hyper-V 2012 for our non production environments. The licensing costs are just too good to pass up. I have also been playing around Hyper-V in my home lab and it is pretty solid. I would say if it all possible and if you have the licensing for it, install SCVMM.
 
That is true, but iirc you CAN upgrade existing VMs to v10 with the native client (I may be wrong, but that seems to be what several community threads have implied at least), which makes absolutely no sense to be able to do if you can't then manage it with the same client. This is the problem where users are running into, and the point where a lot of users are getting lost as they then go looking for the web client they are told they need only to find out that it's not available for stand alone hosts. And they only find out once they find one of the numerous threads on the community because it's so poorly documented and explained by VMware.

Oh wow. It does. I thought the C# would only go to v8 since that's all it lets you create...but if you tell it to upgrade a VM's HW it goes to v10 without asking.
 
Oh wow. It does. I thought the C# would only go to v8 since that's all it lets you create...but if you tell it to upgrade a VM's HW it goes to v10 without asking.

I will file this bug NOW if it isn't filed already - that's complete bullshit and we have to fix that immediately.
 
I completely agree. But the point is it's a pretty crappy precedent to set saying "... all new features will require the Web Client ..." but not inform single host users that the Web Client isn't available, let alone advise them of any caveats as such.

Well, the webclient has never been free - a solution is coming to extend that (much like NetJunkie, I can't talk about it yet), but you shouldn't be able to ~get~ a HW10 VM without it - the fact that the thick client is doing that is a major issue I'll make sure we fix ASAP.
 
Hey Lop i was gonna do that but u beat me to it! i was gonna check to see if it was already known out there due to it being on the communities...
 
If you only have 25 legacy VMs on 4 hosts....I don't see why you wouldn't just consolidate that to 2 or 3 and get like an Essentials license. Unless those are just monster VMs you're way low on consolidation ratio.
Definitely could do better with consolidation, but as I said - that would certainly require some new hardware, not to mention a reasonable SAN upgrade while we're at it.
Proposing "we can replicate exactly what we're doing now, but with minimum $20k in hardware plus $1000 for licensing" probably won't go over so well.

Currently the most powerful host in service is a dual Xeon E5405, and that's hosting 2 old SQL servers that eat up 12GB RAM each (physical machines had 32 each), and another 8 machines that all share the remaining 8GB. I thought ESXi 5.5 would be awesome, since it removes the RAM limit - but I'm holding off for now to see if the Free web client actually comes out.

Anyway, moving around some clients to free up hadware to test out XenServer - thanks again for all the feedback guys.
 
Why would you need an upgraded SAN? And you're not replicating what you have now. You have no HA capability or other more advanced features right now.
 
Why would you need an upgraded SAN? And you're not replicating what you have now. You have no HA capability or other more advanced features right now.
Upgraded SAN for expansion - we're maxed out on what we have right now, and it's a bit of a patchwork. 2 hosts are already on local for a majority of their storage.

As I said in the first post, we don't need HA on the clients that are currently virtualized. They can be down for 24-48 hours without any negative business impact.
Virtualizing the majority of production servers is not possible ATM because we have to comply with vendor specifications (B2B retail). Primary vendors dictate a lot of the systems we run so it can be interfaced back to them for things like JIT stock delivery and ISO compliance, and those software vendors dictate the config they support - none of them support VMs today.
 
Definitely could do better with consolidation, but as I said - that would certainly require some new hardware, not to mention a reasonable SAN upgrade while we're at it.
Proposing "we can replicate exactly what we're doing now, but with minimum $20k in hardware plus $1000 for licensing" probably won't go over so well.

Currently the most powerful host in service is a dual Xeon E5405, and that's hosting 2 old SQL servers that eat up 12GB RAM each (physical machines had 32 each), and another 8 machines that all share the remaining 8GB. I thought ESXi 5.5 would be awesome, since it removes the RAM limit - but I'm holding off for now to see if the Free web client actually comes out.

Anyway, moving around some clients to free up hadware to test out XenServer - thanks again for all the feedback guys.

You don't need the web client. The webclient is only for HW10, which is AHCI disks for guests (limited scope), SMP FT (you don't have licenses), and similar bits all of which are licensed. Now, you shouldn't be able to upgrade to HW 10 with the thick client, but if you can, just avoid doing so for now and 5.5 would do fine for you with the thick client.
 
I'm still confused as to what's causing you to run away screaming from ESXi v5.5 toward XenServer, other than:

1) a bug which allows you to upgrade a VM to v10 and not be able to manage it (which is being addressed above)
2) that they haven't yet released a web management platform independent of vCenter, which from all accounts looks to be forthcoming, and doesn't matter much anyway since the only deltas in lack of C# management capability are related to licensed features

What did I miss?
 
I'm still confused as to what's causing you to run away screaming from ESXi v5.5 toward XenServer, other than:

1) a bug which allows you to upgrade a VM to v10 and not be able to manage it (which is being addressed above)
2) that they haven't yet released a web management platform independent of vCenter, which from all accounts looks to be forthcoming, and doesn't matter much anyway since the only deltas in lack of C# management capability are related to licensed features

What did I miss?
Why do you dodge in boxing when someone is telegraphing a haymaker?

The bigger picture is that I need a longer term infrastructure to get into, and now is the time I can get that going. VMware was the default and easy choice, and if I get any deeper into it it'll just be that much harder to pull out.

The way I see it, their half-baked release of 5.5 isn't a mistake - it's a clear and obvious message that they have no intention of keeping the Free Hypervisor a relevant option for the SMB space. Similar to when they neutered the Backup APIs. I can completely understand the strategy, and the need to protect the revenue stream - it's just not something I can justify the cost on at this time.
 
Back
Top