How Long Do Hosting Platforms Keep Your Logs? (2026 Comparison)
Vercel deletes your runtime logs after 1 hour on the free plan. Netlify gives you a day. Fly's search window is a week. Here's every major platform's log retention in one table, what it means when something breaks, and how to keep logs longer on any of them.
Log retention by platform
| Platform | Free tier | Paid | Keep longer |
|---|---|---|---|
| Vercel | 1 hour | 1 day (Pro) · 3 days (Enterprise) · 30 days with Observability Plus add-on | Log Drains (Pro+) or an external logger |
| Netlify | 1 day | up to 7 days (plan-dependent) | Log Drains (paid plans) |
| Fly.io | live tail only | 7 days (log search in Grafana) | Fly Log Shipper to an external store |
| Render | live tail (default view: last hour) | short native history; HTTP logs on Pro+ | Log Streams to an external provider |
| Railway | — (no free tier; Hobby from $5) | recent logs only (days, plan-dependent) | No log drains — Railway's docs point you to third-party shippers |
| Supabase | 1 day | ~7 days (Pro) · ~28 days (Team) — plan-based | External logs via Log Drain-style exporters or an SDK |
| AWS Lambda (CloudWatch) | — | you configure retention per log group; default never expires — and you pay per GB-month | Subscription filters to Firehose/an external sink |
| Cloudflare Workers | live tail only | 3 days (Workers Logs) | Tail Workers or an external pipeline |
// Verified against official docs where fixed (Vercel, Netlify, Fly, Cloudflare); plan-dependent // figures marked approximate. Providers change these numbers — always confirm on current docs.
Why an hour of logs is not logs
- ▸The bug a user reports tomorrow morning is already gone on Vercel Hobby, Netlify Free, and Supabase Free. You debug from screenshots.
- ▸Nothing to query means nothing to learn: no error-rate trends, no 'when did this start?', no correlation across days.
- ▸Every platform's own docs point to the same answer for real debugging: forward your logs somewhere with real retention. The window exists to protect their storage bill, not your incident response.
- ▸Compliance and support tickets regularly reference events older than every free-tier window on this page.
The 2 ways to keep logs longer
Platform drains & shippers
Vercel Log Drains, Netlify Log Drains, Fly Log Shipper, Render Log Streams, CloudWatch subscription filters. Usually paid-plan-only. You pick a destination (Datadog, Axiom, S3, FlareLog) and pay both sides: the drain and the store.
An SDK inside your app
FlareLog ships logs out as your code runs — works on every plan in the table above, including free tiers where drains aren't offered. 10,000 logs/month and 7-day retention free; 90 days, search, and error alerts on Pro ($19/mo). Plus an MCP server so Cursor or Claude can query your production logs directly.
Logs that outlive the platform window
One SDK covers Vercel, Netlify, Fly, Railway, Render, Cloudflare Workers, and plain Node — structured logs, error alerts, cost burn monitoring, 90-day retention on Pro.
Start free — 10k logs/mo →FAQ
Which hosting platform keeps logs the longest?
AWS CloudWatch technically keeps logs forever by default — but you pay per GB-month for storage, so 'unlimited' becomes an uncontrolled line item. Among the platforms with built-in windows, Supabase Team (~28 days) and Vercel/Enterprise with Observability Plus (30 days, 14-day view window) are the longest. Most platforms keep between 1 hour and 7 days.
Why do platforms delete logs so quickly?
Storage cost and compliance. Storing every log line for every customer is one of the biggest cost centers for hosting providers, so they treat logs as a short-lived debugging aid, not a system of record. That's why nearly every platform offers an escape hatch — drains, shippers, or SDKs — that forwards logs to a dedicated store where you control retention.
How do I keep my logs longer than my platform allows?
Forward them out in real time. The options are the platform's drain/shipper feature (Vercel Log Drains, Netlify Log Drains, Fly Log Shipper, Render Log Streams, CloudWatch subscription filters), or an SDK that ships logs from inside your app. The SDK route works on every plan — including free tiers where drains aren't available — and is what FlareLog does: 7-day retention free, 90 days on Pro.
How long does Cloudflare keep Worker logs?
Workers Logs keeps 3 days on the paid plan; the free plan only has a live tail. FlareLog's Tail Worker captures every invocation — including crashes — and retains logs up to 90 days on Pro.
Is long log retention actually useful, or just hoarding?
It's useful the moment something breaks slowly: an error that happens twice a week, a customer complaint from last Tuesday, a regression introduced by a deploy three weeks ago. Short windows only catch incidents you're watching live. For everything else — which is most incidents — retention is the difference between debugging and guessing.