Log File Storage

Logs are write-once, read-rarely-but-urgently. You generate gigabytes that nobody touches for weeks, then an incident hits and you need last Tuesday's payment-service log in the next two minutes. files.link fits that access pattern: cheap, durable storage for the long tail, with retrieval that's a plain HTTP GET when you actually need it. Shipping logs is curl-friendly by design — no IAM roles, no instance profiles, no SDK to vendor into a slim container.

Your logrotate postrotate hook or a CI step uploads the compressed file with three calls: POST /v1/files/{folderId} for a presigned URL, PUT the gzipped log, POST /v1/files/confirm-upload with the id, authenticating via the raw API key header. A logrotate hook looks like:

postrotate
  F=/var/log/app/app.log.1.gz
  UP=$(curl -s -H "Authorization: $KEY" -d '{"files":[{"name":"app-'$(hostname)'-'$(date +%F)'.log.gz"}]}' https://api.files.link/v1/files/$FOLDER)
  curl -s -X PUT --data-binary @$F "$(echo $UP|jq -r .urls[0].url)"
  curl -s -H "Authorization: $KEY" -d '{"ids":["'$(echo $UP|jq -r .urls[0].id)'"]}' https://api.files.link/v1/files/confirm-upload
endscript

Use folders to keep the archive navigable — partition by service, host, and date so finding payments/2026-06/host-3 is obvious. For logs past their hot window, move them to the cold storage tier: it cuts the per-GB cost substantially while keeping retrieval instant, so a compliance archive you might never read costs little but is still one GET away if an auditor or an incident demands it.

Everything is encrypted at rest, which matters because logs routinely leak more than people expect — emails, tokens, internal IPs, stack traces — so keep log folders private and pull files through signed URLs rather than public links. The honest scope: files.link is log *storage and archival*, not a log *analytics* platform. It doesn't index, tail, or run queries the way Datadog, Loki, or an ELK stack does — pair it with your search tooling for live debugging, and use files.link as the durable, low-cost backing store and long-term archive those tools age data out of. Prepaid credits keep the bill predictable even when log volume spikes during an incident.