Skip to content
Register

One year with CloudCone: the good, the bad, the weird

Reviews by FyreJun 15 replies 2.2K views
FyreJun

Member OP

#1

It is my humble opinion that perhaps I ought to share my experience, and I apologize if this post is overlong. I am not certain anyone will find value in it.

August 2024: 512MB KVM, 1 vCPU, 20GB SSD, $18/year. Disk I/O was 800 MB/s. It was my understanding that this was acceptable.

November 2024: I/O dropped to 400 MB/s. Support suggested it was "noisy neighbor" and would resolve. It did not resolve.

January 2025: Unexpected free upgrade to 1GB RAM, same price. I was surprised. The I/O remained degraded.

March 2025: CPU steal time visible in top. Load averages of 3.0 on a single core. I opened a ticket; it is my humble opinion that perhaps they oversold the node.

June 2025: Network latency to Asia doubled. I did not complain, as the price remained low.

The weird: the free upgrade arrived with no announcement, no changelog. One day it simply was. I do not understand their business model, and I apologize for my ignorance.

Has anyone else experienced this pattern? I am curious whether loyalty is rewarded or punished by such providers.

annexbi

Member

2 #2
FyreJun said:
The free upgrade arrived with no announcement

Ah yes the stealth upgrade (classic CloudCone move (or so I hear (I have never used them (well I tried once (the signup form broke (nested bugs))))) ) they probably needed to shuffle you to a different node (oversold (obviously)) and the extra RAM was cheaper than admitting the old node was dying (which it was (I guarantee it))).

I am on OVHcloud now (worse support (better honesty (about the honesty (they have none)))).

push. done. coffee.
tifiucem

Member

🇹🇷 cheap vps hunter

#3

I am on CloudCone 2 years. Same story they gave me free upgrade too then disk went to 200 mb/s. I pay 15 dollar not 18 cheaper than you xixixi.

But seriously dont expect nothing you get what you pay for. I cant complain for this price. Şu fiyatlara göre iyi!

çok ucuz, çok hızlı
dadlime

Member

stable > testing

#4
tifiucem said:
I pay 15 dollar

The price discrimination is real. They know who will tolerate what (and at what threshold). I have been tracking my I/O since March (dd if=/dev/zero of=test bs=1M count=100). It forms a perfect decay curve. Semicolons help me cope. The data does not.

Loyalty pays in small RAM bumps. It costs in dignity.

Joy

Member

6 #5
dadlime said:
Loyalty pays in small RAM bumps

I am agree lah!!! I am buy CloudCone very cheap xixixi!!! At pirst I am tinking good deal!!! Now I am know better xixixi!!!

But I am stay because I am lazy to migrate xixixi!!!

kate3

Member

#6

The "free upgrade" is a node consolidation play. They moved you to a newer host with higher density and lower per-VM resource allocation. The virtualization tax increased; you received more RAM but less reliable CPU scheduling and degraded disk QoS.

Check your cgroup limits: cat /sys/fs/cgroup/cpu/cpu.cfs_quota_us. If it is -1 with high steal, you are on an oversold OpenVZ-style scheduler even if the control panel says KVM.

I have measured this pattern at three budget providers. The upgrade is never free.

virsh list --all | wc -l: 47
readerkrakw

Member

Backup Paranoid

#7

What kernel is the CloudCone node running, 5.15 or 6.x

tamuma

Member

#8

800 to 200 in a year? Ouch

kate3

Member

#9
readerkrakw said:
What kernel is the CloudCone node running

The kernel version is irrelevant for detecting the scheduler behavior I described. What matters is the cpuacct controller and whether cfs_quota is enforced. On my last CloudCone VM the node was running 5.4.203 but the hypervisor was hiding CPU topology entirely. Cat /proc/cpuinfo showed QEMU Virtual CPU version 2.5+ across all cores with no physical core mapping.

Check your steal time with mpstat -P ALL 1 5. If softirq is elevated alongside steal, you are fighting for disk interrupt time too, not just CPU cycles.

virsh list --all | wc -l: 47
readerkrakw

Member

Backup Paranoid

1 #10
kate3 said:
Cat /sys/fs/cgroup/cpu/cpu.cfs_quota_us

On my CloudCone box that file does not exist. The path is /sys/fs/cgroup/cpu,cpuacct/ on this kernel. And cpu.cfs_quota_us is -1. So you are correct.

I have no backups of this VM. I am testing my recovery plan by not having one. So far I am failing the test.

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft