You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
🐛 fix(web): publish status and usage as one pair (#2002)
The dashboard and the admin status page each loaded the status snapshot and the
usage counters into their own resource, refreshed on its own five-second
interval. Whichever half settled first was written straight to the page, so a
card could show listings counted seconds apart from the accepted-request total
printed above it. Neither combination was ever a state the server held.
Both pages now drive one resource that reads the snapshot and the counters
together, so every published pair comes from a single refresh. A refresh tick
that lands while the previous load is still outstanding is skipped rather than
stacking a second request on a slow endpoint, where the responses could arrive
out of order. A tick can also arrive after its route has been left, since the
interval outlives the page; reading the disposed resource would panic, and there
is nothing left to refresh.
The usage half of the pair carries its own outcome instead of failing the whole
load. `/+stats` is operator-scoped and answers `401` to every other reader, so
pairing that failure with the snapshot would leave anyone without the scope
looking at a blank dashboard, and a first load whose counters fail has no
earlier pair to fall back on. A pair whose usage half failed reports that
failure where the counters would go, never counters measured in another refresh.
0 commit comments