New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
I am not willing to disagree with you, I am just stating the obvious.
never mind bro is here now
Hello everyone,
as it’s (sadly) common in such situations, the information provided by our customer is lacking a few points, which is obviously to make us look bad, or to be precise, worse than we are and than the situation was. Considering our own communication with the customer, I am sure that this is done on purpose which is very sad considering that we are known as a customer-friendly, flexible and fair provider to 99% of the users here. But no problem, we will correct the misinformation:
1) The customer is talking about 453 servers in the beginning so the first users in this thread already came to the assumption we’ve been running 453 VMs on a single host. This is not the case and never was. We only delivered 202 VMs (I don’t know why the customer told a different number) - and guess what, that was to make (asap) sure we get rid of any bottlenecks on the hardware side.
2) We don’t “lie” about dedicated cores and we don’t fraud / scam anyone. Our understanding about dedicated cores is well-known, communicated clearly to everyone who asks, was explained several times on LET and is a de-facto market standard in this price range: Every customer is allowed and usually of course able to fully use all resources of their cores. We will not intervene on high CPU usage, we won’t throttle anything and we won’t kick anyone out for having a lot of servers with this usage pattern. But: This concept works as long as not all customers use 100% of all of their CPU cores 24/7 - and we all know that this is the same with every single provider (!) in this price range. Everyone who’s believing in something else should just configure an AMD EPYC server that would suit as a VPS node, look at the price and do some calculations - the numbers don’t lie and we don’t lie, too. Long story short: Dedicated cores mean you are allowed to use them and you are usually able to do so. Thousands of customers can confirm this is the truth.
3) The policy told in the paragraph before is made sure by monitoring all nodes 24/7/365 regarding a CPU load of less than 50% so CPU spikes of VMs are never an issue. We do live migrations if needed, customer’s won’t even notice that.
4) Dedicated cores are not the same as “pinned” cores. A pinned core is a physical core pinned to exactly one virtual core, never being assigned to other VMs, even when idle. I want to point it out again: No one in our price range does that, it’s not possible. Just forget about it. Prices double or triple with pinned cores.
5) We are a relatively small company having “only” 15 employees in total and a few ten thousand customers. As I already explained to the customer, a single customer ordering 453 VPS more or less at once is very unusual and a single customer ordering 453 VPS and needing all 4 cores of them 24/7 with 100% load (that’s 1812 cores running on full load!) is a situation we never had before after 11 years in the VPS market. When we saw the big order, we set up new hosts and delivered about 202 VPS to the customer, we even set all products to “sold out” to make sure other clients don’t get impacted. The 100% usage of all cores on all VMs simultaneously was not expected before and not communicated to our staff in any way. (If he did so, we would have found a solution in time.) So we just ordered a few more nodes to split the load in the next days right after we noticed this issue (as the new hardware went to unusual load soon). Unfortunately, the customer asked for a full refund without even giving us a day to solve the problem which was reported to us just a day before.
6) Running > 100 VMs on an extremely powerful dual AMD EPYC node is not an issue at all concerning what I pointed out before, as long as you monitor your systems and have enough resources available to migrate peaks away. Bigger companies might run thounsands or ten thousands of physical nodes while we only run hundreds. This will make a difference in such order peaks of course but it doesn't make us scammers.
7) Over the last night, we balanced all VMs to other nodes (partially nodes with way better CPUs than sold in this product category to split the load in the best possible way) and thereby solved the client’s issue. We also told him that but he is not responding to our messages anymore. So I want to make clear that the screenshots provided by the client show a situation that is not intended by our company in any way. The customer of course knows that and still decided to post about a situation solved half a day ago. Why?
8) We have already learned from the situation and will add nodes to the cluster with CPUs which are more optimal for such a workload (more cores, less GHz) and we will (we must) implement a limit of xxx (yet to be determined) VMs per customer.
9) We also offered the customer to switch to several (six) “business dedicated servers” providing him with more than 1500 own physical threads at same price - he refused.
10) Another lie told by the customer is that we “refused to refund” him. The servers were three orders in total. The first was fulfilled, the second one was partially fulfilled but refunded entirely (!) although the customer used the servers for a week and the third one will be refunded entirely, too, although we also delivered some servers of this order. We never claimed something else and I wonder where this information comes from if not from us. We only refuse to issue a full refund for an order that was fulfilled because of an issue which was fixed within 24 hours - especially not after consuming this amount of resources and after we invested lots of time and money (€€€€€) to make the customer happy. Sadly, it didn’t work here and we’ve learned from it.
11) Some competitors fake the steal time by manipulating the host’s kernel. Just do your benchmarks of the real-world performance… We don’t do that. I mean we could, we have great sysadmins, but we don't do it.
12) It’s interesting that he also attacks our label PHP-Friends - which is currently running on a completely different cluster, not being able to tell anything about the label’s products.
13) Doesn’t matter in a technical way but we are talking about customer having a different name on his ID than in his PayPal account, buying nearly 500 VMs and using 100% CPU 24/7. Just think about it, and we will do, too.
Have a great evening
Tim
--
https://www.reddit.com/r/hetzner/comments/10l66i7/does_hetzner_have_disabled_the_cpu_steal/
https://forum.netcup.de/administration-eines-server-vserver/vserver-server-kvm-server/p78405-steal-bei-dedizierte-kerne/#post78405
It it pretty much is if resources are claimed to be dedicated but really aren't. Like someone already stated above the practice is quite likely to run afoul of Germany's "unfair competition" laws. If this isn't worked out i'd actually highly recommend OP to take this to the German consumer protection watchdog. If they agree it would basically make a lawyer bill appear in the providers mailbox accompanied with the request to sign a cease and desist declaration containing a hefty contractual fine.
Yeah it could be, but thankfully laws are interpreted by human beings.
ChatGPT
The message addresses several points regarding a customer's complaints:
Clarifies that they provided 202 VMs, not 453.
Explains their policy on dedicated cores and CPU usage.
Assures monitoring to prevent CPU spikes.
Differentiates dedicated cores from pinned cores.
Describes their response to the unexpected demand and the customer's refund request.
Discusses capacity to handle VMs on powerful nodes.
Explains steps taken to resolve the client's issue and questions the timing of the complaint.
Acknowledges the need for better hardware for certain workloads and plans to implement customer VM limits.
Offered alternative solutions to the customer, which were refused.
Clarifies their refund policy.
Denies manipulating host kernel for performance.
Notes an attack on their label PHP-Friends, which operates separately.
Mentions discrepancies in the customer's identity and usage pattern.
Overall, the message aims to correct misinformation and address the customer's concerns.
Bro, don't talk so much, just do it
Thanks for the extra info. Personally used Avoro in the past and never had an issue, if all is true and extra time could have sorted this, I don't really see much of an issue overall.
This is hilarious. It's just a bunch of excuses. You need to specifically write on the page that the cores are not dedicated. And you never solved anything. They all still had a high steal. I'm not interested in any dedicated servers. If I want ed dedicated servers I would go with the best which is hetzner. If you can't actual provide dedicated cores, then don't advertise it as such and then talk about "pinned cores".
"We only refuse to issue a full refund for an order that was fulfilled because of an issue which was fixed within 24 hours".
What are you even on about? The steal was high before you cancelled the servers today and yesterday. The screenshots I have are from yesterday. And I ordered the servers on the 19th and 20th. Nothing was fixed.
You have not refunded a lot of the UNDELIVERED servers. You have not refunded the third one and ALL of them should be refunded because these are NOT dedicated cores.
The name on the Paypal is my business partner. That has nothing to do with this. I attacked php friends and dataforest because you're all under the same company.
It's crazy that you still refuse to refund even after admitting that the cores are not dedicated.
Limiting the amount of orders in bulk/per account should be the least if they know they can't handle the sales.
Or just limit the amount of VPSs that someone can buy on each plan so they don't have to build new nodes to handle the load?
Their VPS features "instant setup", all I will say.
How should i do it? I'm not the concerned party here... but well, if you feel so safe that you feel the need to taunt me i might offer OP to take his stuff to the Verbraucherzentrale. You are advertising something you don't deliver giving you an unfair edge over your competition. That's basically the definition of "unlauterer Wettbewerb" and you probably know what "kostenpflichtige Abmahnung" is. It's the stuff that can hit you even if you just word your Impressum the wrong way.
Fully agree, but this was something unexpected for them and hasn't ever happened, I'm sure they can in the future.
Issues happen and if a resolution was offered within a quick timeframe, that's more than you get from 90% of other hosts within the industry.
Are you kidding me? I have hundreds of screenshots of the servers all having an extremely high steal. They didn't fix anything.
But is what they say true regarding them trying to help and find a solution? Within the time they state? Or is that made up?
Dude. I, as a customer do not care if you pin cores, shuffle them around, use some crazy advanced AI (Hi Luca!) to balance or whatever hosting blackmagic you do, but
dedicatedis fuckingdedicated. If you advertisededicatedthen I should be able to use 100% of CPU, 100% of the time, no matter what my neighbors are doing. And by use I mean use, not wait for that CPU to be free akasteal.Also this is not a topic about what other people/company does. YOU advertise
dedicated, not them.--
Can you imagine offers with
dedicatedIPv4 and you can [only] use them if other customers shutdown theirs machine because it'sconcept works as long as not all customers use 100% of all of their IPv4.Like what the actual fuck?
You just allow
burst, you have nothingdedicated.Dedicated cores don’t guarantee that your YABS score won’t drop. The problem is that CPUs get slower when there’s more load, even if there’s CPU cores remaining. Moreover, if Netcup allocates vCPU cores, there’s a chance that CPU steal will present past 50% host node CPU load due to hyperthreading.
On an AMD EPYC 7702 for example, the CPU score in a VM will drop by almost half if the host node is loaded to 50%. Despite the host node having 64 vCPU cores left, the YABS score drops because the CPU can’t maintain as high of clocks.
Take a look at my testing here: http://paste.frocdn.com/qayabesevo.apache
Chill out man, if all providers who advertise dedicated reserve 100% of the resources, we would have to pay 2x or 3x the prices for such machines.
Try order 500VMs from anyone else.
I won't deny that it's not economically viable to have dedicated cores / threads at this price point.
However, when I'm sold dedicated cores / threads, I assume that these are resources reserved for me. And I don't understand that this only means that I'm entitled to generate a high CPU load, but that CPU power isn't guaranteed. And I don't understand how this is different from shared cpu cores, except that you're “non-contractually” committing to better load balancing to make it work on a small scale.
I find this a misleading business practice, and it's very specific to German hosting providers.
Again, I don't deny that competitors do the same thing.
But it seems to me to be a dishonest practice, especially for foreign competitors. Most of whom don't use this kind of commercial practice.
Well, on the product page
That is clearly misrepresentation to say the least, if cpu core performance is not consistent. I agree that disk access speed is limited by shared I/O as hardware resource is shared, but with respect to cpu cores, dedicated word implies that the resource is exclusively allocated to this VPS
Not only that but pinned core will mean practically nothing to like 99.99% of all customers as the term is virtually unused around hosting.
Glad we're finally on the same page.
No, because they are dedicated. We explained why the issue happened with your order, we apologized for it and we fixed it. I mean, you still claim that we didn't, but everyone can see your screenshots are more than a day old. Personally I assume that you didn't even check it again as you opened PayPal disputes a few hours after my answer included in your first post.
They will be refunded if they aren't already. Since our PayPal account is always paid out automatically on Mondays and PayPal holds back the money from your disputes (created after the automatic payout to our bank account), one refund is simply pending but initiated.
They aren't even providing a full refund for false advertising. The guy was lying in the beginning when speaking to him but later admitted that the servers are oversold. I'm not interested in a business dedicated server. I want what I paid for not oversold servers. They never fixed anything.
Here is the full conversation:
No their resolution was to buy "business dedicated servers". They never fixed anything. The steal was always high.
Yes, you can do it, of course you know "einstweilige Verfügung" and "Schutzschrift" and "Hauptsacheverfahren" - and if we would loose, also netcup, Hetzner and all other providers are not allowed anymore to offer "dedicated Cores" on virtual servers, that would cause a large wave of "kostenpflichtige Abmahnungen" between providers. Do you also get so upset when a provider offers “unlimited fair use traffic” and limits you to 10 MBit/s after 1 TB?
They are NOT dedicated. The steal is higher than any VPS I've ever used. You are lying to people selling an oversold server in the hopes they don't actually use it. Stop with the jargon about "pinned cores". If someone can't use the cores fully, it is not dedicated. It's shared and actually 10x worse than a VPS in this case.
I want a full refund. That is only fair for this false advertising and lies. And remove the "dedicated" on the root server pages.
PS: For transparency, I'm also willing to share some vitality parameters of the nodes in the Avoro KVM cluster. Nothing fraudy here
And one of our webservers, running on the same cluster as our customers: https://imgur.com/a/07YtvxT
Hm, somehow I can't see the 100% steal here.
@dataforest
for any product with dedicated cpu cores, my only question is
are cpu cores exclusively dedicated to the VPS?
Thanks, the screenshots do help as they didn't exactly say they would come to a resolution in the tickets as they said they did in their post just now..
You can't advertise dedicated when it isn't. Even your terms and conditions don't say anything about that. What are you not understanding here?
Fucking exactly!
If I am on a hunt for
dedicatedcores because of my load type then I want to finddedicatedcores, even if this cost 2x or 3x.I don't want to go each provider and ask support what
dedicatedmeans because some providers "cheats" - I could understand that when summ3rh0st2025.com havededicatedcores I should be alarmed, but Dataforest was kinda trusted provider that should not go to such tactics.Why 500 VMs? I ordered ONE
dedicatedVM, but provider decided to cram additional 452 VMs on my node and I am getting rekt for that because those neighbors decided to use theirsdedicatedCPU thread/vCore? Yeah, no thanks.