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
Thank you for your advice.
I only monitor my dedicated servers with node exporter, which run the most of the "real workloads".
I've never really monitored those i had in ph or other small providers, because i know the usage is very small only.
It seems like you are not getting it at this point:
A yabs benchmark doesn't cause "bursts", a yabs benchmark causes high sustained cpu utilization on one thread and on all threads of a machine for a couple of minutes each.
On a node full of Blockchain users it could appear like there is still plenty of space for customers while this is completely not the case.
The reason behind that is that those blockchain applications need to interact with transactions. It also does that while you believe that this is only causing a static load of eg 20%. But every now and then there are very small CPU bursts that almost utilize all cores because of things happening in the chain.
Another issue with those blockchains is that they mostly use CPU instructions like avx512 which causes very bad performance for other guest as the infinity fabric of genoa CPUs isn't able to handle this kind of extreme use very well (in a fair style).
You can easily achieve that also while only having 4% cpu load on your machine.
You will rather need tools like perf and pidstat to analyze whats going on exactly.
Do you want me to start a detailed excessive analysis session on the exact blockchain you had been running because of what, a 4,95€ transaction being at stake?
This will not happen unfortunately, please choose between the two options that i provided just recently.
I don't really get this insane blockchain obsession people have with us anyway.
If i would make money of it i certainly wouldn't use promo offers with shared resources for that.
That's not a proxmox issue, the reported metrics are the metrics of all processes eating resources within a VMs cgroup.
We eg decided to spawn a dedicated iothread for every disk to improve io performance while guest having high cpu load, this way you can actually utilize 500% cpu on a hypervisor while only having 400% in cpu cores.
This is also the reason why we display more memory usage then reported in VMs.
Most people don't know that linux is using memory for caching.
We just report what your VM is doing on our host.
There is no logic in our web backend to connect to the exact os you have deployed in your vm to then retrieve the exact resource utilization that your os reports inside your vm to you.
Hopefully this problem can be resolved in a good way
the answer is simple, please refund me.
enough with all the charade and reasoning.
That makes a ton of sense why you'd do this and that is certainly reflected in the I/O metrics I see in Netdata. That dedicated iothread is something the guest would never see.
Also I know Netdata reports 60% total consumed RAM but I have about 8GB cached because unused RAM is wasted RAM right?
You can use Netdata Cloud with a few free agents or even export what Netdata sees into prometheus. Then you just run a local prometheus instance at home or where ever and you just store that data.
Thanks for choosing one of the provided options, please state that in your ticket
Ticket#207124
Looks like you've got your stuff together
You may want to take a look at https://cortexmetrics.io/ to run prometheus at scale. We are in the process of setting it up so that we can easily handle per-minute exact billing by monitoring all the metrics in our openstack cloud, which we will be migrating to soon btw.
That looks neat. I'll add that to my to research and maybe implement. Also yep... self hosted for almost a decade. Started with Nehalem Xeons to Ivy Bridge i3s and at home I run a ML110 G9(with various VMs for specific workloads/tasks) and an EliteBook for Gitlab CI/CD runner.
Then I have my 2 VPS nodes for production and connectivity (the downside of living in NZ, latency is terrible but I have amazing download and upload).
One in the states for a pyphon mod bot, and then one with you as a game server
.
All of this is round the clock monitored, locked down hard and tweaked to the fit the environment.
Oh that's actually funny.
I get why you aren't sharing your metrics but instead accuse us of anything that comes up to your mind...
I just issued your refunds and now hereby exclude you permanently from doing business with us again.
Thank you for the intense experience, wishing you all the best
Why would i share my metric when you're not?
and thanks anyway.
please don't ever enable crypto payment option when you're not even supporting it
We started our journey with those ibm softlayer dual E5-2680v2 Supermicro hosts back in 2018
Oh don't be bothered by what you are reading here.
The worst that could happen with your game server is a never ending loop of a script trying to start the game over and over again always resulting in a OOM.
Happens quite often actually
I believe 1-2x a week
We already accept crypto payment manually for years now.
I don't get how a payment method influences the acceptable use policy of a service.
As your account got blocked you still should be able to click on the screenshot in the mail our system sent you.
You can post it if you like, but we don't think a lot of doing something like that without consent.
I am not sure if i am still interested to answer your message but feel free to express yourself
@ProHosting24 I am also your customer, I use it for RDP for general browsing & codding stuffs as my internet connection is slow. But if you suspend servers even without any notice or warning that is bit worries me. When we open chrome CPU go up sometimes. Not sure what your demon does, but i would recommend you enable a warning system so that client would know they need to limit their usage or should move on to different provider or need upgrade their vps or many be move to different plan, and after warning if client don't respond or stop, you are free to suspend it without any refund also.
We don't do auto suspension based on CPU load, this only affects blockchain workloads regardless of the load as the impact is not quite measurrabe.
What you just described is our normal way of operating up to this point. But we are going to change that as manual intervention starts to get to time consuming at this scale.
I already thought of suggesting something like a automail on high load with the user having the option to choose an upgrade or to respond by the intent of lowering his load.
Of cause without any interaction that involves humans on our part.
And if this workflow doesn't cover the last 5% of those affected users then we are still happy to interact with them manually.
Blockchain shouldn't be done on shared resources to begin with...
Hell training LLMs shouldn't be done on shared resources either...
When restock we need blackfriday
Is there any plan to restock the Frankfurt VPS with a monthly payment of 4.95 euros?
@ProHosting24
Is encoding of multimedia files allowed?
This results in 100% CPU usage for a long time.
I ask... because I read absurd things...
I purchase a service, and you must be able to offer it, if you offer those resources, you must always guarantee them, otherwise don't make these packages...
I have already come to compromises on the disk and the monthly bandwidth. (yes... it says SSD the purchase and then you find an HDD with a speed of 80mb/s on good days, it says 2gbps and on good days there is 1gbps).
But knowing that now you are blocking people without warning, without reaching the deadline and without a justifiable reason (because I repeat YOU are offering those services and you must guarantee them even if they are 100% 24 hours a day).
There are people who use it for work.
You shameless people! This offer doesn’t say anything about dedicated cores. There are plenty of other providers offering VDS for that—why not use them instead of abusing this one?
greed. simple as that.
while I agree that using the advertised ressources has to be possible, with virtualization one has to keep in mind, that the borders between vservers will never be as hard as for two seperate bare metal servers.
things are getting more and more complex and as already has been pointed out above the label "100% CPU usage" does not cut it. for a lot of workloads sustained full cpu usage will usually not be a problem. it's rather the specific cases, where apps or software demands a certain specific part of the CPU architecture, like cache or function set (AVX mentioned above). obviously this also has to be shared but in standard virtualization environments it's up to the hypervisor to handle it. and there is usually no dedicated split in that because it does not help with the performance for the bigger amount of clients with more average usage patterns. imagine you start regulating cache for instance and assigning a hard limit of only few KB to each guest instead of allowing accessing more, when needed and others do not in the same second.
if 20 out of 200 guests already fight for as much as possible from that precious cache everyone will suffer. and yet you won't get to high usage percentages in any view. instead you'll see wait cycles or steal or whatever. high load yes, high CPU percentage no.
if you hard limit that, the 20 users with the specific workload would still complain, because their apps would run miserable. and the performance for the other 180 would still be bad too, because it's counteractive to the intention of that caching.
as said, it's not that simple anymore. but from a providers point of view it is only natural to cut off the rather specific group with specific use patterns and keep the larger group with more balanced workloads and be done with it.
You can see from what he wrote that he doesn‘t have much knowledge overall, maybe he doesn’t even have bad intentions.
But yea beside those blockchain abusers, users like him pulling resources 24/7 like that ist just crazy. Most of those users explained so far that they are recording some weird livestreams from multiple individual streamers to save them 🤷♂️
I get why some people occasionally do that for their jellyfin, but dudes like him are the reason why we are in need of more automation.
The worst thing about this is that normal users start to get paranoid about geting suspended and stuff
It doesn't say anything about dedicated cores, nor that they are shared.
But you know how it is if a person buys and reads 4 core 3.1ghz, and sees shared... it becomes a useless, unsellable product, it wouldn't even start.
Come on...
The resources of that server on which we have been placed is already shared.
No farm sells a dedicated server with such basic resources.
On that server dedicated, if you have done the calculations well, you know how many users you can insert.
But if you, the seller, want to be smart by offering more than you can actually offer, that's your problem, not the customer's.
The customer is purchasing a virtual machine with those resources to operate with software that can occupy those resources offered... If I buy a vps it is because perhaps I cannot operate some things on my personal computer, otherwise what is the point of a vps?
Not everyone has to host a website with 3 pages in html...
A VPS has to do operations and the operations can sometimes be complex and take a long time (depending on the performance of the machine) if the machine is underperforming it is obvious that the operations take time.
I was giving the example for encoding some multimedia content, if this takes 2 hours and I have a list of one hundred contents, this can take several days, but it is not prohibited by the terms and services, so what? do I get banned?
I hope so. Get a VDS.
VPS is always shared resources and sensible use.
And have you been suspended?
Obviously not.
Can you be considered ungrateful?
Yes, totally.
Feel free to cancel your service with us and request a refund as you are so unhappy
What do you want with a 5€ VPS ? You have to know whats the diff between VPS, VDS and dedicated server before buying
I'm writing here because I want to stay calm.
For now I can work, but I don't want to be blocked because the rules are made at the moment.
Proposals for economical VDS are accepted.