Struct PollHealth 

pub struct PollHealth {
    pub feeds_tracked: i64,
    pub polled_last_hour: i64,
    pub overdue: i64,
    pub last_poll_secs_ago: Option<i64>,
    pub oldest_poll_secs_ago: Option<i64>,
    pub never_polled: i64,
    pub in_backoff: i64,
    pub badly_broken: i64,
    pub failure_kinds: Vec<(String, i64)>,
}

Aggregate poll health, for the public stats page.

Deliberately aggregate-only. No user counts, no error rates, no per-feed detail: this is published to anyone, and a reader does not need to know how many people use an instance or which feeds are failing. What it does answer is the only question the page exists for — is the poller keeping up?

Fields

feeds_tracked: i64

Distinct feeds the poller is responsible for.

polled_last_hour: i64

How many were polled within the last hour.

overdue: i64

Feeds whose next_poll has passed — the backlog. A healthy instance clears this every tick; a growing number is the signal that the poller cannot keep up with the feed count.

last_poll_secs_ago: Option<i64>

Seconds since the most recent poll of any feed. None before the first.

oldest_poll_secs_ago: Option<i64>

Seconds since the LEAST recently polled feed was polled — the worst staleness any reader is currently seeing.

None when any feed has NEVER been polled, because that is a worse staleness than any finite age and reporting the finite one would make the page read healthiest exactly when it is least healthy.

never_polled: i64

How many feeds have never been polled at all.

in_backoff: i64

Feeds currently in error backoff (consecutive_errors > 0).

One of the two states that stop feeds updating, and previously visible nowhere: consecutive_errors was written by bump_feed_errors and read by nothing outside the backoff calculation — no page, no endpoint. Worse, a feed in backoff is NOT counted in overdue, because backoff is applied by pushing next_poll forward. So the one number a reader might have checked moved the wrong way: a feed failing every fetch made overdue look BETTER.

badly_broken: i64

Of those, how many have reached [BADLY_BROKEN_ERRORS] consecutive failures — retried 2h40m apart rather than every 5 minutes.

Not “will not recover on their own”: the backoff ceiling is 24h at ten errors, and any of these recovers on its next successful poll. See [BADLY_BROKEN_ERRORS].

failure_kinds: Vec<(String, i64)>

Failing feeds grouped by cause, descending, as (kind, count) — fetch, status, body, parse.

Counts, never identities. /stats is public and states that it reports machines rather than people: no per-feed detail, never which feed and never whose. A cause histogram keeps that promise and still answers the question badly_broken could not — whether sixty feeds are failing for sixty reasons or for one. Had this existed, #159 would have read fetch: 60 on a page anyone could load, instead of costing a production investigation.

Trait Implementations

impl Clone for PollHealth

fn clone(&self) -> PollHealth

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for PollHealth

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Eq for PollHealth

impl PartialEq for PollHealth

fn eq(&self, other: &PollHealth) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for PollHealth