@feigi/fleet-ctl 3.47.0 → 3.47.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/package.json +1 -1
- package/skills/run-team/SKILL.md +22 -9
package/package.json
CHANGED
package/skills/run-team/SKILL.md
CHANGED
|
@@ -2185,9 +2185,13 @@ while :; do # one tick
|
|
|
2185
2185
|
cap_out=
|
|
2186
2186
|
fi
|
|
2187
2187
|
# gh pr list is GraphQL, so $rl still holds. The pass is budgeted at 4 core
|
|
2188
|
-
# REST units per listed PR on top of the 200 floor
|
|
2189
|
-
#
|
|
2190
|
-
#
|
|
2188
|
+
# REST units per listed PR on top of the 200 floor. The 4 is the measured
|
|
2189
|
+
# ceiling of one ci-state.mjs call on the counter `gh api rate_limit`
|
|
2190
|
+
# reports, for single-page reads: 2 on the common path, 4 for a PR whose
|
|
2191
|
+
# head dropped required jobs.
|
|
2192
|
+
# The workflow and run lookups bill a second core bucket that `rate_limit`
|
|
2193
|
+
# does not report and this gate does not guard. A shortfall pauses this
|
|
2194
|
+
# tick's pass.
|
|
2191
2195
|
if [ "$rl" -lt $((200 + 4 * count)) ]; then
|
|
2192
2196
|
if [ -z "$budget_out" ]; then
|
|
2193
2197
|
echo "WATCHER DEGRADED: core REST budget=$rl < $((200 + 4 * count)) needed for $count PRs — CI polling paused, silence is NOT green"
|
|
@@ -2283,12 +2287,21 @@ DEGRADED line names the rest as unwatched until a pass comes back under the cap.
|
|
|
2283
2287
|
|
|
2284
2288
|
**The budget is gated twice, and only the second gate announces recovery.** The
|
|
2285
2289
|
tick-top floor (`200`) covers the probes themselves. Once the open-PR list is
|
|
2286
|
-
read, the pass gate reserves `200 + 4 × count` core REST units —
|
|
2287
|
-
|
|
2288
|
-
|
|
2289
|
-
|
|
2290
|
-
|
|
2291
|
-
|
|
2290
|
+
read, the pass gate reserves `200 + 4 × count` core REST units — 4 per listed PR
|
|
2291
|
+
on top of the floor, so the reserve grows with the list instead of staying a
|
|
2292
|
+
flat 200 that a long list drains mid-pass. The 4 is the measured ceiling of one
|
|
2293
|
+
`ci-state.mjs` call's cost on the counter `gh api rate_limit` reports (measured
|
|
2294
|
+
2026-10-13): 2 on the common path, 4 when the head dropped required jobs, for up
|
|
2295
|
+
to 100 jobs and 100 rules — more pages cost more. That counter is not the only
|
|
2296
|
+
core bucket one call bills. The lookups `actions/workflows`,
|
|
2297
|
+
`actions/workflows/<id>`, `actions/workflows/<id>/runs` and `actions/runs/<id>`
|
|
2298
|
+
bill a second one that `rate_limit` does not report and the gate does not
|
|
2299
|
+
guard; the boundary is per route, so `actions/runs/<id>/jobs` is on the gated
|
|
2300
|
+
counter. Every response says `X-Ratelimit-Resource: core`, so only the reset
|
|
2301
|
+
epoch and the used sequence tell the two apart. An exhausted unreported bucket
|
|
2302
|
+
shows up only as per-PR `rate-limited` verdicts, which the blind-PR path
|
|
2303
|
+
already latches. A shortfall at either gate pauses the **whole** pass through
|
|
2304
|
+
the one budget latch and never polls a prefix of the list: a prefix runs
|
|
2292
2305
|
in list order, so the same tail PRs would go unread every tick with no line at
|
|
2293
2306
|
all (never polled, so nothing marks them blind), while the PRs that were polled
|
|
2294
2307
|
would run dry mid-pass and each emit a blind line. The DEGRADED line from the
|