1732 posts tagged with "Self-Hosting"
Running your own PaaS and infrastructure on machines you own
Not Us, But Also Not Not Us: Fly.io's July 22 DFW Outage and the Upstream You'll Never Meet
On July 22, 2026, emergency maintenance by an upstream network provider cut connectivity for Fly.io apps in Dallas for about thirty minutes. Nobody at fault worked at Fly.io — and no tenant could route around it. Counting the hops between your deploy and the physical plant shows what hosted status pages can't tell you, and what changes when you own the hardware.
Fly.io's Bill Is Now Six Meters With a 2-Hour Preview: Pricing the Predictability Cost Against One Flat Hetzner Box
Fly.io metered volume snapshots in January 2026 and Managed Postgres inter-region traffic in February, while new accounts get a 2-hour trial. A worked six-line monthly bill for a typical small app — quiet month vs busy month — against one flat Hetzner CX22 with backups.
Fly.io Started Charging for Traffic That Used to Be Free: What the Two New 2026 Meters Cost Against Flat Hetzner Bandwidth
Fly.io metered volume snapshots in January 2026 and Managed Postgres inter-region traffic in February. A worked line-item bill for a two-region app under the new meters against a flat Hetzner pair, plus the sensitivity cases — bigger data, heavier replication, costlier regions — where the meters multiply.
The Free Tier Is Dead (Except Twice): A September 2026 Audit of Where a Side Project Still Lives for $0
Only Render and PythonAnywhere still offer free-forever hosting in September 2026 — Railway and Fly.io are trial-credit-only and Heroku has been paid-only since 2022. A verdict table for all six providers plus worked per-project math showing why one €3.99 Hetzner box split across side projects beats every meter.
H100 Rentals Crashed 70%. Should You Still Buy the GPU?
H100 rentals fell from $8-10/hour in late 2024 to about $2.50 on specialized clouds. A worked own-vs-rent comparison shows the breakeven moved from 27% to 87% utilization — ownership is now a bet you justify with measurements, not the default.
Heroku Killed 12 TLS Ciphers: The Legacy-Client Audit Your Self-Hosted PaaS Needs Too
Heroku removed 12 legacy TLS cipher suites and enabled TLS 1.3 with HTTP/2 on August 24, 2026. Here is which clients break, a reproducible openssl/sslyze/testssl.sh audit for your own domains, and the cert-manager plus Gateway API baseline to match it.
Kubernetes 1.37 Kills the Last Idle Pod: Native HPA Scale-to-Zero vs KEDA for Sleeping Apps
Kubernetes 1.37 graduates HPA scale-to-zero to beta and enables it by default: queue workers can shed every replica with minReplicas 0 and no KEDA. HTTP apps still need a buffering layer, and beta isn't ready for a tenant SLA — the head-to-head, the idle-bill math, and the checklist.
MCP Has 10,000 Servers but Three Production Gaps: Identity Propagation, Timeout Budgets, and Structured Errors for Deploy Agents
Ten thousand MCP servers and 97 million monthly SDK downloads made connecting agents to tools trivial — but an agent that can deploy and roll back production needs scoped identity, timeout budgets, and machine-readable errors first. This post designs that deploy surface, tool by tool.
One GPU Per Sandbox: What Modal's Whole-Card Bet Costs Against E2B and Daytona's CPU-First Lanes
Modal puts whole GPUs inside sandboxes while E2B stays CPU-only and Daytona's GPUs are ephemeral-only. The reason is physical — Firecracker and gVisor cannot pass a GPU through — and the billing consequence is an 8x cost swing on GPU busy fraction alone.