Can FailSafe keep serving stale entries beyond StaleWhileRevalidateDuration when the factory consistently throws?

Question

FusionCache v2.x implements a stale‑while‑revalidate strategy that returns immediately stale data while an asynchronous refresh runs. The StaleWhileRevalidateDuration controls how long the stale entry remains valid after expiration. When the factory delegate throws during a background refresh, the library logs the error and preserves the stale entry until the duration lapses. If the factory keeps failing, the FailSafe feature extends the window by FailSafeMaxDuration, after which the entry is evicted and subsequent requests propagate the exception. This raises several design questions:

Does FailSafe allow stale data to be served beyond StaleWhileRevalidateDuration when the factory consistently throws?
How does the distributed lock used for the L2 background refresh influence stale serving across multiple nodes?
Is there a circuit‑breaker or throttling mechanism that prevents a hammering pattern during the FailSafe window?

Tasadduq Burney · Answer

Direct Answer
Yes. When the factory delegate throws during a background refresh, FailSafe keeps the stale entry available for up to FailSafeMaxDuration after the original StaleWhileRevalidateDuration window expires. The total maximum staleness is the sum of both durations.
Confirmed Facts

StaleWhileRevalidateDuration defines the normal async‑refresh window after TTL expiry.
FailSafeMaxDuration is additive and only activates when a background refresh has already failed.
If the factory succeeds at any point during the FailSafe window, the entry is refreshed and normal stale‑while‑revalidate behavior resumes.
FailSafe applies only to entries that are already stale (past TTL) and have suffered a failed background refresh; fresh entries are unaffected.

Steps to Verify This Behavior

Configure a cache entry with a short TTL (e.g., 1 s), StaleWhileRevalidateDuration (2 s), and FailSafeMaxDuration (10 s).
Provide a factory delegate that always throws an exception.
Call GetOrDefault repeatedly and observe that stale data is returned for roughly TTL + StaleWhileRevalidateDuration + FailSafeMaxDuration (≈13 s) before the exception propagates.
Check the logs for “FailSafe” entries to confirm the extension is active.

Can FailSafe keep serving stale entries beyond StaleWhileRevalidateDuration when the factory consistently throws?

1 answer

Direct Answer

Confirmed Facts

Steps to Verify This Behavior

Your perspective belongs here.

0 question comments