Member
@nate_pad IMO 512MB might launch a very small world, YMMV heavily depending on player count and render distance. Take it with a grain of salt, but PaperMC with aggressive flags can squeeze into 400MB. You will be swapping, though.
I ran a two-player server on Leaseweb's 512MB plan once. Worked fine. Not recommended for public, but possible.
Mas tipo 😂
SPF
Reputação
Nossa mano
Member
Runs alpine + wireguard + pihole + a python script that scrapes HostHatch stock and texts me when the $2.50 plan drops.
It actually works.
#!/bin/ash
curl -s https://HostHatch.com/cheap | grep "Out of Stock" || sendmail ...The script has never fired. I check manually every day. Sunk cost, I guess.
Ronwitactually I am running on KnownHost 128MB VPS... very small(´・ω・｀)
Basically it is just irssi + screen + oidentd... 24/7 bouncer for basically 6 years no reboot actually
Once I tried weechat but basically OOM killer came for me... back to irssi128MB is basically lifestyle
Member
- 64GB Ampere Altra
- 2TB NVMe
- 10Gbps unmetered
- Customer usage: 0.003% CPU, 4KB RAM, 0.00 Mbps
As said, we did not suspend him. We promoted his tweet. Our CAC dropped 40% that quarter.
Dry humor: the HTML file is now our careers page background.
Member
Marketing.
Not ridiculous.
The ridiculous:
Me.
512MB.
Prague.
Running.
Nothing.
Uptime.
847 days.
Why.
I.
Forgot.
The.
Root.
Password.
Member
That's not ridiculous, that's just OpenVZ with a bad cgroup limit and no systemd watchdog.
Real ridiculous: I watched a "KVM" provider on RackNerd whose "virtualization tax" was so heavy the guest kernel couldn't even see its own /proc/cpuinfo flags. Customer thought they had AES-NI. They had QEMU emulation. Six months of "why is my WireGuard slow" before they checked.
At least your 512MB is honest. Probably.
Aiyah 847 days lor can just use single user mode what
Unless lah the provider lock console then cham liao
I got one GreenCloudVPS NAT box 128MB run docker already 3 years leh
docker in docker some more
Can
Regarding the Forgotten Root Password Phenomenon and Related Infrastructure Pathologies
As I mentioned above in my analysis of resource utilization patterns (though this thread has moved rapidly), the intersection of uptime obsession and credential hygiene presents a fascinating case study. I propose the following taxonomy of "running nothing" deployments:
1. The Accidental Museum
A. Server provisioned for purpose X
B. Purpose X abandoned
C. Uptime becomes purpose
D. Password lost due to:
I. Password manager migration
Ii. Offboarding of original admin
Iii. Simply... forgetting (see @Prague)
2. The Deliberate Artifact
A. Intentional minimalism as aesthetic
B. Cost below notice threshold
C. Social media cachet ("look at my uptime")
3. The Provider-Locked Zombie
A. Console access paywalled
B. Reboot means death
C. 847 days and counting
Table of contents for my upcoming post: swap usage on "idle" 64GB instances, the HostHatch $2.50 plan arbitrage economy, and why Alpine in 128MB is actually cheating (musl libc, no systemd).
Bro I ran llama.cpp 7B q4_0 on 512MB shared vps once
Not recommended lol
Took 40 minutes to reply "hello"
The ridiculous part was I was paying more in overage than just renting a 3090 on OVHcloud
Ngl the math is simple: 7B at q4 = ~4GB VRAM, or ~4GB RAM + swap death
512MB can hold the tokenizer and like 2 layers before it starts murdering your disk
@Chen docker in docker on 128MB is actually more impressive ngl