Testing SkySilk
-
@black3dynamite said in Testing SkySilk:
@brandon220 said in Testing SkySilk:
What do you all think of them using Proxmox as the base virtualization platform? I have no experience with it but it does look appealing. Thoughts?
I do like Proxmox Web Interface.
Looks nice, but the underlying product and the company behind it....
-
@bnrstnr said in Testing SkySilk:
@scottalanmiller said in Testing SkySilk:
@brandon220 said in Testing SkySilk:
What do you all think of them using Proxmox as the base virtualization platform? I have no experience with it but it does look appealing. Thoughts?
ProxMox is a joke as it is. Using it as their base is ridiculous. Given that they are a pure Linux provider, using KVM at all doesn't make much sense. They should be only on LXC, I would think.
That is literally enough for me to no longer consider them a serious provider in any way.
https://help.skysilk.com/support/discussions/topics/9000041880
Our current infrastructure is based on LXC containers, which are served via ProxMox virtualization. Eventually, our goal is to be able to offer multiple types of virtualization such as KVM and others, but at the time of this writing our main focus and product offerings are centered around ProxMox.That doesn't make sense. ProxMox is a management layer, not virtualization. ProxMox' purpose is to do KVM and LXC transparently. But mixing the two on the same hardware is nutty unless you have only one box in your entire environment. And even then, nutty, but not AS nutty.
So their logic makes no sense. They have to have KVM in place to have ProxMox. So their statement is either a bold faced lie, or an admission that they have zero clue what they are doing.
-
There is no reason to have ProxMox if you are doing LXC. LXC has great interfaces as it is. ProxMox would actually just make things slower, more fragile, and more difficult to extend.
-
That sounds lazy of them to use Proxmox for LXC. Why not two separate servers, one for LXC and the other for KVM?
-
@black3dynamite said in Testing SkySilk:
That sounds lazy of them to use Proxmox for LXC. Why not two separate servers, one for LXC and the other for KVM?
Exactly. They have to build a cloud interface on top either way. ProxMox doesn't appear to be adding anything of value.
My guess is, like Cloud@Cost, they are using a third party product because they don't understand the moving parts involved and think that they can quickly get up and running with the Jurassic Park Effect and hope that things don't fall apart on them. Sounds, to me, like they built a product that they don't know how to support or how it actually works. All development, no operations.
-
Are they using Proxmox Interface or a custom one on top of Proxmox?
-
@black3dynamite said in Testing SkySilk:
Are they using Proxmox Interface or a custom one on top of Proxmox?
Has to be custom, no cloud in ProxMox.
-
"Ceph is a trademark of Red Hat, Inc. Proxmox is a trademark of InkTank Storage, Inc. "
Actually InkTank Storage is long gone. They made CEPH and were bought by Red Hat. InkTank and Red Hat have nothing to do with ProxMox. RedHat is American, ProxMox is Austrian.
-
We have @Matt_SkySilk around here now. Maybe he has some insight or information for us.
Just noticed that he had joined the community. -
@aaronstuder said in Testing SkySilk:
My first VPS took about 1 minute to come online. Running benchmarks now.
Currently testing the Basic Nano plan.
How did your benchmarks come out? What did you use? GeekBench?
-
@scottalanmiller Benchmark Failed. Discontinued testing. You can't reload OS.
-
@aaronstuder said in Testing SkySilk:
@scottalanmiller Benchmark Failed. Discontinued testing. You can't reload OS.
Can't reload? In what sense?
-
-
@aaronstuder said in Testing SkySilk:
@scottalanmiller said in Testing SkySilk:
Can't reload? In what sense?
In any sense.
Like, if anything goes wrong with it, you are just hosed?
-
Reload OS? Meaning you can't reinstall?
-
@wirestyle22 said in Testing SkySilk:
Reload OS? Meaning you can't reinstall?
Apparently, that's pretty rough. I'm guessing that there is no ISO loading, which I didn't notice, so likely there is not. Given that they are on LXC, that would be hard to do. You have to rely on them to provide good, up to date images. Which is why them not being up to date is a bigger deal. Not that you can't update from say Fedora 27 to 28, but it's not fast, nor a good process. So restoring a simple server on there that would be seconds on another platform might take an hour or more and manual steps!
-
Hi everyone, Matt from SkySilk here! (apologies for the long post, please bear with me, I promise future responses will be shorter) I wanted to make sure everyone’s comments are addressed…I’ll post a TLDR and then the expanded response below.
@aaronstuder thanks for checking us out and sharing our link here! We take all user feedback to heart and enjoy seeing what people have to say about our new platform.
@scottalanmiller Thanks for your feedback, glad you enjoy the option to choose your own processor!
TLDR;
- At this time we only offer LXC containers for better machine performance, we have a whole host of different features and products in the works.
- Please contact me if you would like to get a trial account to run some benchmarks I will be more than happy to assist.
- openSUSE - This distro was the least popular out of all templates provided during our beta, so as you can imagine it’s not at the top of our list. Over 20,000 VPS spun off and this template was less than 1%. That said we like feedback and will try to facilitate individual requests whenever possible.
- Ceph is way more robust than using a Raid10 in a larger environment (in our experience), which is why we’ve opted to go that route.
- While a true OS re-install doesn't exist yet (we're working on it), you can workaround that by immediately taking a snapshot after deploying and using that as a way to reset your VPS to a clean state if you don't want to delete/re-deploy.
Expanded Response:
True, we only offer OpenSuse Leap and it is two versions old - we have decided to focus on the crowd favorites (or at least based on the feedback from our beta users) which have been Ubuntu and CentOS. We are working hard towards providing a wider range of distros to choose from, as well as turn-key and custom applications. That said, we do appreciate feedback and can prioritize certain distros if anyone makes a request.
I am glad that you noticed we use Ceph in lieu of RAID. We use Ceph for our clustered storage backend with a triple replicated RBD pool. Unlike RAID, this means your block storage device is online even if a node falls down (or gets shot by the other nodes[1]). While the performance may not be the same as RAID10, we feel that data safety and high-availability take precedence.
On to your next point, all software and technology comes with trials and tribulations, that said, Proxmox is most definitely worthy of being the backbone of SkySilk. It provides full cluster management, failover capabilities, a rather strict enforcement of secured multi-tenancy[2], and an expansive API upon which we have been able to develop a robust and stable cloud platform. When you have multiple clusters in different regions, a management layer that recovers from a node fault in the middle of the night is a godsend. Plus it isn’t particularly bloated and doesn’t require a license to press the power button.
By the way, there is no harm in running a composite LXC + KVM environment on the same hardware, we just have to be diligent with resource provisioning so that we can guarantee quality of service. That said, KVM and custom ISO offerings are on the docket, and will most likely be served on better hardware. And hopefully Windows as well. For now though, we only offer LXC because the reduced overhead of containers allows us to squeeze just a little bit more value out of our hardware. Interestingly, this also translates into performance benefits for our users at the expense of being locked into a specific kernel version.
True, we do not currently support the ability to re-install an OS but it is coming on our future features list. As a workaround, users can always take a snapshot and use that to revert to a clean OS state. And as referenced above, if you are in need of something specific as far as images or distros go we encourage you to reach out so we know what to prioritize.
As for the stability of our product and our ability to support it, we have just finished an extensive beta testing period that saw over 20,000 LXC container deployments by selected beta testers. We have thoroughly vetted & tested our platform to the point where it is now a quality, production-ready product we are extremely proud of and understand how to support well.
That’s not to say we don’t have areas where we can improve; we know that’s the case without a doubt, and are working hard to grow our team and platform at the same time. In that same vein, our approach up to this point has been anything but lazy, and we don’t plan on changing that any time soon.
Footnotes:
[1] https://en.wikipedia.org/wiki/STONITH
[2] If you can’t get your software to run feel free to reach out to support, we do grant more leash for certain use cases.
