Skrime suspended me for "high utilization"

forest · October 5

@HostDZire said:
Now we know he doesn't just use high bandwidth, he also does 100% cpu

LET hosts be aware, he might just cost you 1000$ by paying 2-3$

Just waiting to find out that his average packet size is 50 bytes and he's doing 500k PPS too.

Obelous · October 5

@forest said:

@HostDZire said:
Now we know he doesn't just use high bandwidth, he also does 100% cpu

LET hosts be aware, he might just cost you 1000$ by paying 2-3$

Just waiting to find out that his average packet size is 50 bytes and he's doing 500k PPS too.

And slamming the disks until they start clicking

ascicode · October 5

@HostDZire said:
Now we know he doesn't just use high bandwidth, he also does 100% cpu

LET hosts be aware, he might just cost you 1000$ by paying 2-3$

Your chunk ist not interested, look on your own problems.

@Alyx said:

@forest said:

@SKRIME said:

@forest said:

@SKRIME said: @Alyx Fair point to raise, but the suspension had nothing to do with traffic. That part is genuinely unmetered and stays that way. It was CPU.

Be careful saying it's genuinely unmetered. OP is the person who used 800 TB in a month on a $3 VPS from Cloudblast and is the reason why they no longer sell unmetered and instead limit traffic to a mere 5 TB/month.

Why suspend if CPU usage is excessive rather than throttle? If you throttle to 25%, the server will show 75% steal but will still be operational. Then you've removed the immediate stability threat and have time to do the question-asking.

@forest Fair question. Throttling would keep the node stable, but it doesn't solve the actual problem. A VM that needs its cores at 100% around the clock is simply on the wrong plan, and a VM throttled down to 75% steal isn't really usable for the customer either. So we suspend, talk to the customer, and unsuspend once we know what's running and how it will be handled, which is exactly what happened here.

The reason throttling may be better is that, if combined with an automated email, it gives the customer the opportunity to rectify it themselves. Sometimes a customer might just not realize how much CPU they're using.

A notification ahead of any action is generally something I would recommend.
This way the customer can take care of the issue without you / the provider introducing unneeded downtime for the customer.

Like, I personally would be a little annoyed if my provider shut down my service, just because some process got stuck.

First the customer should read emails daily, but some only react, if a service is down.

HostDZire · October 5

@ascicode said:

@HostDZire said:
Now we know he doesn't just use high bandwidth, he also does 100% cpu

LET hosts be aware, he might just cost you 1000$ by paying 2-3$

Your chunk ist not interested, look on your own problems.

This kind of shit is also my problem, and I am taking interest in these because they are related to providers problems. You are a client, so you wouldn't understand . however other providers can easily understand whats wrong happening here.

JohnFilch123 · October 5

My Skrime server has been suspended as well for high cpu usage but ticket helped and I adjusted the load, so running great now.

@HostDZire said: Now we know he doesn't just use high bandwidth, he also does 100% cpu

Sometimes those are related. High bandwidth usage in certain cases leads to higher CPU load.

zed · October 5

@xHosts said: This is leading me more and more to do LET validation of orders and refusing orders from people who act like this who will instantly create a thread before allowing anytime at all for a review of the issue

you absolutely should.

xHosts · October 5

@zed said:

@xHosts said: This is leading me more and more to do LET validation of orders and refusing orders from people who act like this who will instantly create a thread before allowing anytime at all for a review of the issue

you absolutely should.

If someone contacts me, tells me they need a server with a,b and c and the budget I will aim to work towards that (if its realistic), I have customers from LET who push a fair amount of bandwidth and use a fair amount of CPU but balance things as much as possible while also in the past 6 months been able to pick up some killer hardware 60+ physical cores, near 1TB of DDR4 ram which when a customer needs this type of power it can be done.

The only thing I wont fall for is spending time working out something fair then being told X provider will do this, asking me to match it but I know within the month they will have some type of issue, at that point I withdraw my special deal and wont provide to them in the future.

I am a fair person, but wont be messed around for just crumbs of income when I could easily assign them resources to another customer who is a local business happy to pay x5 that and will not bring hassle with them

averagedatahoarder · October 5

My goal was not to shame Skrime in fact I am/was quite a happy customer of theirs, I was just a bit annoyed that they didn’t give me any warning resulting in downtime for all my services.

I would’ve appreciated throttling or a few hours warning at least instead of instant suspension.

That being said, the suspension was reasonable and I’ll keep my CPU load down in the future

averagedatahoarder · October 5

@SKRIME said:
@xHosts Pretty much at the same time. The ticket and the thread came in within moments of each other.

@forest Fair point for the case where someone simply doesn't realise. In our experience though, a VM sitting at 100% permanently is most often one that's been compromised, usually through a weak password, and then suspending straight away is the safer move for everyone, the customer included. We're around most of the time, so the wait afterwards tends to be short, and we've handled it this way since the start, always together with the customer.

In this case I was running a one-time import job overnight. My VPS wasn't compromised, etc. Fair assumption though I guess, I just would've appreciated some notice beforehand y'know? Caused a bit of downtime

JohnFilch123 · October 5

@averagedatahoarder said: some notice beforehand

I suspect you were sleeping at night, which makes it a bit challenging.

averagedatahoarder · October 5

@JohnFilch123 said:

@averagedatahoarder said: some notice beforehand

I suspect you were sleeping at night, which makes it a bit challenging.

12 hours, throttle to 50% steal

buggedout · October 5

@averagedatahoarder said:

@JohnFilch123 said:

@averagedatahoarder said: some notice beforehand

I suspect you were sleeping at night, which makes it a bit challenging.

12 hours, throttle to 50% steal

Then title would have been: Skrime throttled me for "high utilization"

averagedatahoarder · October 5

@buggedout said:

@averagedatahoarder said:

@JohnFilch123 said:

@averagedatahoarder said: some notice beforehand

I suspect you were sleeping at night, which makes it a bit challenging.

12 hours, throttle to 50% steal

Then title would have been: Skrime throttled me for "high utilization"

Or perhaps there would be no post at all :P

CloudHopper · October 5

@averagedatahoarder how do you burn through 800TB bandwidth in a month? Not judging, just genuinely interested how you manage to blast a port that hard.

Edit: the truth is I'm asking because I will probably need to move ~30TB of data as soon as possible at some point so I'm wondering how you manage to max the throughput so much

averagedatahoarder · October 5

@CloudHopper said:
@averagedatahoarder how do you burn through 800TB bandwidth in a month? Not judging, just genuinely interested how you manage to blast a port that hard.

Uhhh erm well

rpqu · October 5

@averagedatahoarder said:

@CloudHopper said:
@averagedatahoarder how do you burn through 800TB bandwidth in a month? Not judging, just genuinely interested how you manage to blast a port that hard.

Uhhh erm well

It's a CDN, right?

averagedatahoarder · October 5

@rpqu said:

@averagedatahoarder said:

@CloudHopper said:
@averagedatahoarder how do you burn through 800TB bandwidth in a month? Not judging, just genuinely interested how you manage to blast a port that hard.

Uhhh erm well

It's a CDN, right?

No

averagedatahoarder · October 5

Much of the usage came from me having to mirror a 273TB dataset from SFTP server A to SFTP server B.

Also, proxying content and hosting bandwidth-intensive APIs/tools. All traffic was organic.

barbaros · October 5

@averagedatahoarder said:
Much of the usage came from me having to mirror a 273TB dataset from SFTP server A to SFTP server B.

Also, proxying content and hosting bandwidth-intensive APIs/tools. All traffic was organic.

And you thought that much usage is OK in a shared VPS environment?

averagedatahoarder · October 5

@barbaros said:

@averagedatahoarder said:
Much of the usage came from me having to mirror a 273TB dataset from SFTP server A to SFTP server B.

Also, proxying content and hosting bandwidth-intensive APIs/tools. All traffic was organic.

And you thought that much usage is OK in a shared VPS environment?

I mean... it was marketed as "10Gbps unlimited" ¯\_(ツ)_/¯

CloudHopper · October 5

@averagedatahoarder said:
Much of the usage came from me having to mirror a 273TB dataset from SFTP server A to SFTP server B.

Also, proxying content and hosting bandwidth-intensive APIs/tools. All traffic was organic.

But how did you max out a shared 10Gbps link like that? It's impressive and I'm more interested in the mechanics than the ethics or the reasons.

averagedatahoarder · October 5

@CloudHopper said:

@averagedatahoarder said:
Much of the usage came from me having to mirror a 273TB dataset from SFTP server A to SFTP server B.

Also, proxying content and hosting bandwidth-intensive APIs/tools. All traffic was organic.

But how did you max out a shared 10Gbps link like that? It's impressive and I'm more interested in the mechanics than the ethics or the reasons.

Not that hard to max it out if you know what you're doing (I don't know what I'm doing but eh)

rcy026 · October 5

@averagedatahoarder said:
Much of the usage came from me having to mirror a 273TB dataset from SFTP server A to SFTP server B.

You have a vps with 273TB storage??
What does something like that cost you? I'm guessing it's not one of their standard plans.

averagedatahoarder · October 5

@rcy026 said:

@averagedatahoarder said:
Much of the usage came from me having to mirror a 273TB dataset from SFTP server A to SFTP server B.

You have a vps with 273TB storage??
What does something like that cost you? I'm guessing it's not one of their standard plans.

No.

zed · October 6

@rcy026 said: SFTP server A to SFTP server B

OpaqueRegistrant · October 6

@SKRIME said:
@Alyx Fair point to raise, but the suspension had nothing to do with traffic. That part is genuinely unmetered and stays that way. It was CPU.

On our regular plans the cores are shared with the other VMs on the node, so running them pinned at 100% around the clock isn't what those plans are for. That's the same CPU fair use we've described publicly in our offer threads. In practice we only step in when a VM sits at 100% permanently, and that's what happened here.

AI wrote this comment.
Why not give them a 25% reservation or something?

averagedatahoarder · October 6

@zed said:

@rcy026 said: SFTP server A to SFTP server B

Not my SFTP servers!

SKRIME · October 6

@OpaqueRegistrant said:

@SKRIME said:
@Alyx Fair point to raise, but the suspension had nothing to do with traffic. That part is genuinely unmetered and stays that way. It was CPU.

On our regular plans the cores are shared with the other VMs on the node, so running them pinned at 100% around the clock isn't what those plans are for. That's the same CPU fair use we've described publicly in our offer threads. In practice we only step in when a VM sits at 100% permanently, and that's what happened here.

AI wrote this comment.

Why not give them a 25% reservation or something?

@OpaqueRegistrant English isn't our first language, so we use tools to help put what we want to say into words. The content is ours.

On throttling, we've answered that a few times further up in this thread already.

DK87 · October 6

@HostDZire said: This kind of shit is also my problem, and I am taking interest in these because they are related to providers problems. You are a client, so you wouldn't understand . however other providers can easily understand whats wrong happening here.

For me, as a hosting\vps customer - very important for not to seat with such abusers on one node.

Whole story is a joke. Troll bought 6 euros vps and utilized it like a 600 euros dedi, nice.

Skrime did all right.

Unlimited budget vps is always subject of fair usage.

If you're poor - that's not the reason for abuse everyone on your node.

TimboJones · October 7

@averagedatahoarder said:

@s0n1c said:
Based on your thread history of hunting for cheap unmetered bandwidth VPS's and your LET username, are you sure you didnt use like 100tb in less than a month and then pull the "idk whats going on, my usage was completely normal" card? 💀

Idk man, it seems like everyone hates me I have no idea why 🥺🥺🥺

Lacking asshole awareness.

Howdy, Stranger!

Categories

In this Discussion

Skrime suspended me for "high utilization"

Comments

Howdy, Stranger!

Quick Links

Categories

In this Discussion

Skrime suspended me for "high utilization"

Comments