Roadmap to KVM Templates

[ejc317]

I'd like to understand the roadmap to offering KVM templates. In order for proxmox to be truly viable as a commercial solution rather than a cool albeit robust garage/shed solution to virtualization, KVM templating is key.

Solutions like SolusVM and Virtualizor offer this but do not offer HA. I can appreciate the team really focusing on building the groundwork for HA, etc before templating but in this day and age, not offering this, I believe, is a mistake. This is a critical piece that's missing.

I guess ultimately it depends on where the Proxmox team seems themselves in the cloud value chain. Whether it is cloud orchestration, vps administration, etc.


Cloud orchestration platforms such as cloudstack, etc all have the feature with varying degrees of flexibility and functionality but the ability to build upon it is there. I think as other programs seek to work with proxmox (NOCPS, whmcs, hostbill), the team should at least look to this and we are happy to work to develop this as well.

Currently, after 3 months of testing, in order for us to begin deploying this, we need templating and the only solution that is client friendly seems to be dumping Virtualizor on top of a VM inside Proxmox (so nested virtualization) which is anything but ideal.

I would love to see where the team is at on this and if the rest of the community can pitch in to get something like this deployed so that Proxmox can elevate to the next level as a viable commercial cloud orchestration platform as it is very robust in a number of other aspects

 

[dietmar]

I'd like to understand the roadmap to offering KVM templates. In order for proxmox to be truly viable as a commercial solution rather than a cool albeit robust garage/shed solution to virtualization, KVM templating is key.

We have OpenVZ based templates for many years. KVM based templates are of less interest, because MS/RHEL license forbid to make templates.

 

[ejc317]

KVM templates is a key selling point for guys like OnApp, cloudstack, etc -

 

[ejc317]

FYI - the issue with OpenVZ templates is if you're saying people should really be using KVM on proxmox and with san - since our focus is on a truely HA solution, then you need to offer templating - otherwise it defeats the purpose no?

 

[dietmar]

since our focus is on a truely HA solution, then you need to offer templating - otherwise it defeats the purpose no?

HA is totally unrelated to templates.

 

[ejc317]

Did I say it was? I said your focus was on HA and not on templating but maybe its time to focus on that. Look you guys can ignore what I say - but just letting you know some feedback is all. We're going to move on to a more suitable cloud virtualization platform for the commercial side - I ultimately do not know your plans as to the future of proxmox - but for internal use, proxmox is perfectly fine

 

[dietmar]

Did I say it was? I said your focus was on HA and not on templating but maybe its time to focus on that. Look you guys can ignore what I say

I do not ignore what you say. Instead, I try to understand what you say and answer. I just want to point out that we already have a full featured template system for OpenVZ.

but just letting you know some feedback is all.

thanks.

 

[ejc317]

I do not ignore what you say. Instead, I try to understand what you say and answer. I just want to point out that we already have a full featured template system for OpenVZ.



thanks.

You're missing the point. OpenVZ can't use iSCSI SAN, therefore not ideal for a fast, HA setup. openVZ is paravirtualization, not ideal for lots of deployment. So in order to compete with cloudstack, onapp, etc - you need a true baremetal hypervisor with templating capabilities

and to stick our head in the sand and ignore the competition is not sustainable but like I said don't know what your goals are. For what it is - its a good product but can be improved is all I'm saying

 

[tom]

Did I say it was? I said your focus was on HA and not on templating

this is not true. we provide templates since 2008, we also included TKL templates. HA since 2012. and just for the records, you can run OpenVZ on NFS, enabling a true HA solution with containers. OpenVZ containers (and Parallels Virtuozzo, based on OpenVZ) are the global market leader in providing hosted VPS - seems you missed that.

if you want KVM templates, create the "template", backup and restore as many times you need it. search the forum, discussed many times already.

but maybe its time to focus on that. Look you guys can ignore what I say - but just letting you know some feedback is all. We're going to move on to a more suitable cloud virtualization platform for the commercial side - I ultimately do not know your plans as to the future of proxmox - but for internal use, proxmox is perfectly fine

we do not ignore what you say but I correct your postings. and generally speaking, its not the best start to name our software a "garage solution" just because you miss a single feature.

instead, think what you can do to make it happen - join the dev team, pay someone else to develop features. try to understand the open source development process.

just a small hint: open source does not mean that someone works for zero to create YOUR business.

 

[ejc317]

Wow - no one said work for free. I merely said can we understand the roadmap as in where it is in your queue. No one said work for free - you guys seem to be quite sensitive to that.

openvz and others may have been the leaders in virtualization but let me assure you, vmware, etc are the new leaders.

In terms of the small hosted vm universe which I believe is your target - (no large enterprise currently deploys proxmox ...) - then you need to get with the times and create a product that people can use to run their business on.

this is why vps.net runs onapp and tons of other people are using it. Frankly speaking you guys have a great product but being stubborn and refusing to acknowledge the times doesn't get you anywhere. Just because OVH gives people proxmox to use as part of their budget dedicated servers doesn't mean you guys should be content

this is a probust piece of software that can be used to deploy solid private cloud solutions. Again, the title merely said roadmap to KVM templates. No one said do anything for free. And fyi - people run businsesses so to garner a positive ROE, will deploy capital as needed.

If we run our business on proxmox, rest assured you will be paid support (we've spent 3 months testing the software for a reason) but to hide behind the excuse of people have been happy with it blah blah is not an excuse. And frankly would I rather pay a developer to develop something that may be out of date when you release the next upgrade or pay onapp and use their solution? I think the answer is pretty clear.

Don't get defensive when people point out clear flaws - if you don't believe me let's do a poll - how many people on this forum / of your user base would like kvm templates. And FYI - if you had that feature I guarantee you the target audience would be greater.

Finally, with regards to your point on NFS, etc. NFS is way slower than iscsi. Any business worth their salt will use redundant iscsi SANs. Multipathing over 10GBe fiber to iscsi SAN just can't be compared to NFS.

Anyway, seems the answer is "wer'e happy doing what we're doing and attacking potential customers and will stick our head in the sand the second someone brings up other valuable features that competitors have" - we are happy with that roadmap - just means we should look elsewhere

Unfortunate that such a great piece of software can be marred by this type of attitude - and again I reiterate - NO ONE SAID WORK FOR FREE - learn to take some suggestions

 

[ejc317]

And I must say - when I say "garage solution" that is not a typo, It doesn't mean garbage. I mean the majority of people here are not running it in the same capacity as other comparable products. FYI, the goal of the post was to understand the roadmap

a) We're happy to contribute time and money to develop features
b) We're happy to join the dev team

I've been in the hosting industry for 9 years and have seen lots of crappy products and frankly this is one of the better things - i'm just letting you know from a consumer stand point what would really garner you some extra brownie points.

If you don't believe me - just ask - how many people would like kvm templates. If you had that, you'd completely eat OnApp / Solus (not a true cloud orchestrator) and virtualizor's lunch

 

[dietmar]

I really do not understand what you want. I just said 'KVM templates are of less interest'.

So I you want to help to get that implemented feel free to contribute.

 

[dietmar]

a) We're happy to contribute time and money to develop features
b) We're happy to join the dev team

Also see: http://pve.proxmox.com/wiki/Developer_Documentation

 

[spirit]

Hi,
no sure it's related, but I'm working on a kvm vm cloning feature.
(So you can duplicate an existing vm that you have created before). Maybe this can feed your need ?

 

[tom]

we introduce new features day by day. your statements like "we are attacking potential customers" is just very weird.
if you want to contribute something, you know how to do. so do it or don´t do it. if you want to become our customer, do it.

this thread is already moved out of the topic so I will stop answering here.

 

[bpvarsity]

I really do not understand what you want. I just said 'KVM templates are of less interest'.

So I you want to help to get that implemented feel free to contribute.

I to and many others i know would really like KVM templates. I would like to use proxmox over solus and other virtualization software but we would really like to have templates. Not having the ability to use them is a big turn off for a bunch of people. Ive been on many forums and more and more are people asking how to use template with proxmox and why they dont offer it. I think it should be more of a interest.

 

[spirit]

I to and many others i know would really like KVM templates. I would like to use proxmox over solus and other virtualization software but we would really like to have templates. Not having the ability to use them is a big turn off for a bunch of people. Ive been on many forums and more and more are people asking how to use template with proxmox and why they dont offer it. I think it should be more of a interest.

Hi, are you talking about templates like for openvz. (prebuild apps).

Or just vm cloning ? (you make yourself a base vm then you duplicate it).

Because I'm working on this for the next proxmox release.

 

[mmenaz]

Just curious, but what is a "kvm template" compared to just backup a "base" KVM vm and restore with a different ID number, change mac address / hostname / sshkey / ipaddress?

 

[mo_]

I am only reiterating what the staff has already said here, but I felt like it could use repackaging in different wording:

- If you want zero effort templates for linux machines: you use openvz. to connect it to HA you use NFS storage - any other storage solution is probably very low priority, but conceivable...
- If you want zero effort templates for windows machines: youre out of luck since obviously youre not allowed to distribute pre-installed windows installations due to licensing.

- Literally the only thing KVM templates would be good for are net/free/openBSD (-based appliances). Those however only occupy a very small niche and probably havent even been requested once. However - using FreeNAS (FreeBSD based NAS appliance) on a physical server is probably your best bet to get a ZFS based nfs going.

You can however create a windows template for KVM yourself (since YOU do have M$ licenses apparantly) in two different ways with different feature sets:

1) For SAN/iscsi storage you can backup+restore any KVM machine to duplicate it and even keep "backups" of machines, which are essentially checkpoints.
The downside for this is that is uses a lot of storage.

This can all be done in the proxmox webinterface even (if for some reason you dislike the CLI)

Its a trade-off: you get the best io performance for the cost of much higher storage consumption

2) For nfs storage you can use qcow2 images, which feature CoW (copy on write). So essentially you can install a... say windows server 2008r2, start it, patch it, shut it down and never run it again.
After that is done you can create new KVM machines that use a qcow2 image, which uses the previously mentioned VM as its base image. Using this you get a fresh VM, which uses less than 10MB of storage (before you start customizing it). Using qcow2 base images and CoW you can also create all sorts of nested base images and checkpoints and save on a lot of storage space.

These qcow2 images however offer (slightly?) lower io performance than raw images (method 1).

This option is lot more flexible and offers a much higher flexibility and potentially saves lots of storage capacity (if done right). However this can not be done from the webinterface, you have to use the CLI for it. You could create a new VM from the webinterface, not outright start it, and then ssh into the server and replace the newly create qcow2 with a CoW from your base image.

PS: please dont use "cloud", where you actually mean "clustering" since "cloud" is more a bullshit bingo term (pardon my french) used by managers and to excite layman consumers.

 

[spirit]

1) For SAN/iscsi storage you can backup+restore any KVM machine to duplicate it and even keep "backups" of machines, which are essentially checkpoints.
The downside for this is that is uses a lot of storage.

This can all be done in the proxmox webinterface even (if for some reason you dislike the CLI)

Its a trade-off: you get the best io performance for the cost of much higher storage consumption

2) For nfs storage you can use qcow2 images, which feature CoW (copy on write). So essentially you can install a... say windows server 2008r2, start it, patch it, shut it down and never run it again.
After that is done you can create new KVM machines that use a qcow2 image, which uses the previously mentioned VM as its base image. Using this you get a fresh VM, which uses less than 10MB of storage (before you start customizing it). Using qcow2 base images and CoW you can also create all sorts of nested base images and checkpoints and save on a lot of storage space.

These qcow2 images however offer (slightly?) lower io performance than raw images (method 1).

This option is lot more flexible and offers a much higher flexibility and potentially saves lots of storage capacity (if done right). However this can not be done from the webinterface, you have to use the CLI for it. You could create a new VM from the webinterface, not outright start it, and then ssh into the server and replace the newly create qcow2 with a CoW from your base image.

PS: please dont use "cloud", where you actually mean "clustering" since "cloud" is more a bullshit bingo term (pardon my french) used by managers and to excite layman consumers.

Hi Mo,
I'm working on both for next proxmox release.
(vm copy and vm "clone" aka linked clone in vmware world), with simple right-click clone on a source vm without need to backup/restore.

for 2) I have also implemented rbd,sheepdog and nexenta (zfs) cloning features.