her.stnewthe library is open ->
her.st eye and mask logo

hosting

For seven years this machine has stayed up all but eleven hours and fifteen minutes - better than five nines, on metal I also live on. It is not a company and there is no SLA; when something breaks, I am the first to notice, because my own stuff runs right next to yours. It just moved to AMD Zen4.

>99.999%
uptime since 01/01/2019 - 11h15m of total downtime across seven years, on self-hosted AMD Zen4 metal. run by one person, not a business.
what you can run
recommended
lxc

lean system containers - the best perf-per-dollar thanks to near-zero overhead, and RAM, disk and CPU scale live without a reboot. this is what most people should pick.

from $1.00 / shared core
full vm
kvm

full virtualization, your own kernel, max isolation. only if you need a custom kernel or LXC won't run it.

from $3.75 / dedicated core
managed
docker

hand me a docker-compose.yml and I run it. resources scale on demand, usually without reboots.

from $1.50 / shared core
sites
web

SFTP into your webroot with your SSH key. I handle TLS on nginx via Let's Encrypt. no rsync or ftp.

from $1.00 / mo
rate card - per unit, monthly USD
configurable KVM LXC Docker Web*
base price - - - $1.00
CPU (dedicated) $3.75 $3.75 $4.00 -
CPU (shared) $1.00 $1.00 $1.50 -
RAM (GB) $1.25 $1.00 $1.50 -
NVMe (GB) $0.15 $0.15 $0.20 $1.00*
IPv4 $3.00 $3.00 - -
IPv6 $0.00 $0.00 - -
network 1Gbit 1Gbit 1Gbit 1Gbit
DDoS protection 1Tbit 1Tbit 1Tbit 1Tbit

Web*: SFTP (scp, lftp) to your webroot with an SSH key. No rsync or ftp. TLS on nginx via Let's Encrypt. $1 plan = 1GB NVMe. Preferably paid yearly. Network is 1Gbit fair use - 1TB/mo per $5 spent ($20 buys 5TB). LXC backups on request at extra cost. Prefer LXC over KVM unless you need a custom kernel.

chronology - downtime since 2019
2020
11/05 - nameserver
0h09m - maintenance; *.her.st didn't resolve briefly, rest fine
01/07 - everything
5h12m - power surge killed the rack PSU; replaced
2022
28/10 - everything
0h15m - moved to a bigger server, staggered per-VM, no data loss
29/10 - everything
0h02m - IPv6 NIC assignment failed; fixed with a reboot
2023
19/01 - dns / reverse proxy
0h04m - docker overlay FS full, reclaimed space
2024
28/10 - dns / reverse proxy
5h33m - host moved DC; box up, proxy and VPN down
2026
19/06 - everything
~2h00m - migrated the host to AMD Zen4 CPUs; VMs/containers restored as-is, no data loss

not counted in the total: two quarterly node reboots for updates, under 3 minutes each.

how to order
  1. configure - add up the units in the rate card for the plan you want.
  2. message - reach me on XMPP: PM or the public MUC.
  3. pay - Monero (XMR) only. I'll send the address in PM.
the fine print

Your data, your problem. You are responsible for your own data; I am not liable for loss.

Your security, your problem. If your box gets popped, I nuke it. Secure it - keys only, no password SSH.

Don't run Tor exits/relays (one already runs here) or torrents (Germany watches).

Don't do flooding, mass-scanning, spamming, or view/hit botting.

Don't serve copyrighted media or pornography.

If complaint letters arrive about something severe, I nuke the server. I respect your freedom and privacy - but I am not your legal shield.

<- cd ../