Member OP
Still running WireGuard on a 128MB BuyVM slice from 2019. Before anyone asks: no logs, GDPR article 5(1)(c) data minimization, wiped monthly. Surprised the thing hasn't hit article 13 content-filter nightmares yet.
Who's actually pushing lower in 2024? Not "I could if I wanted to" — actually doing it.
Member
Thanks for raising this — it's a fun benchmark for efficiency.
At Contabo our smallest current plan is 512MB, but I checked our legacy inventory:
We do have one customer on a grandfathered 128MB node running a static site. I can DM details if you're curious — no marketing pitch, just impressed it persists.
— Jane @ Contabo
— memory threshold fires every Debian unattended-upgrade. Status page shows "degraded" for 90 seconds, then clears. I stopped waking up for it.
Your WireGuard stable? Mine OOM-killed twice during apt.
Small correction: that node's hosted on HostHatch, not GreenCloudVPS. We migrated the monitoring fleet in March. The graphs don't lie, but the dashboard labels do. :)
Carry on. — Admin
Different nightmare. My 128MB HostHatch Zurich node (grandfathered, $3/mo flash sale) runs OpenWrt x86 in a VM. Luci interface takes 47 seconds to load. But it routes.
Member OP
Stable enough. I compile from source with -Os, stripped, no wg-quick. Static binary under 400KB. The trick is not letting systemd anywhere near it — OpenRC or nothing.
apt needs 64MB just for the resolver now. I do dpkg --unpack manually on a 512MB tmpfs mount, then sync. Painful but OOM-free since March.
Respect. Though in Zurich we have a stranger case: a 96MB HostHatch VM running bird2 for a small ASN. Two full tables would die, but default-only with a static route to the upstream? 34MB resident. Uptime 287 days.
BGP on 96MB. That's the floor, not 128.
Member
I'd love to see that config. Not for replication — we get enough "why won't my 1GB node hold 2 full tables" tickets — but genuinely curious what you stripped.
As a Contabo customer I see the opposite pressure. Customers provisioning 4GB for WordPress because one plugin vendor "recommends" 2GB minimum. The bloat is winning.
128MB for WireGuard is luxury. I run dnscrypt-proxy + unbound on a 64MB BuyVM slice, Las Vegas. Blocklist refresh is the only time it sweats — I cron that for 4AM and accept the 2-minute swap death.
$2/month. My coffee costs more per hour.
Which block storage tier? I tried their 256GB on the 128MB plan and the i/o wait during rsync made it unusable. Fine for append-only logs, terrible for anything random.
