Skip to main content

What Fibe recovers on its own

Not every red moment is a problem. Fibe treats your environments as a resilience contract: transient blips are absorbed and healed automatically, so a momentary hiccup doesn't drag you in. Here's what clears on its own — and what actually needs you.

Heals itself — just wait

  • A brief outage (under a couple of minutes) is ignored. If it persists, Fibe automatically redeploys the environment to bring it back. To avoid thrashing, automatic redeploys hold off for the first 3 minutes after an environment is created and for 10 minutes after each redeploy — detection keeps running the whole time; only the action waits.
  • A reachable app stays viewable even if a single healthcheck is missed. A missed healthcheck never stops or errors a running environment on its own.
  • "TLS pending" / "HTTPS provisioning" is a normal state, not a failure — a certificate is being issued. It clears on its own, usually within a few minutes of a service first being exposed.
  • A stuck launch is picked back up automatically — a launch that sits "in progress" for 30 minutes is re-queued. A temporary infrastructure wobble — a brief network or host-connectivity blip — is retried, not treated as a hard failure.
  • A hung build won't spin forever — a build still running after 45 minutes is failed with a clear timeout error rather than left in "building" limbo.
  • A Genie that momentarily can't be reached is left viewable and quietly redeployed if the outage holds; a mid-provisioning SSL handshake shows as "pending," never an error.
  • Trick runs are the exception: once a run finishes or fails it stays that way — Fibe never redeploys a job run on its own. Re-run the Trick instead. (A launch that stalls before the run gets going is still picked back up automatically.)

Needs you — act

These are real and won't fix themselves:

  • A service that crashes on startup — bad image, missing variable, wrong bind address. Check the logs.
  • A validation error on a Template or Playspec — the message names the fix.
  • A genuinely down or unreachable host — firewall, disk full, key changed. The Marquee's connection test surfaces it.
  • A Trick that never exits — the watched service isn't finishing.

See Common problems & fixes for the exact message → smallest-fix table.

One caveat: funding

Automatic recovery, redeploys, and auto-expiration only run while the Marquee is funded. An unfunded Marquee's environments are left exactly as they are until you fund it again — nothing self-heals in the meantime. See Billing — When your balance runs low.

See also