Public: anyone can read this; the crew can change it.
The edge half of #9622: a person runs 500/s for 60 s through the edge and past it on a restored copy of production, and keeps both files #12135
Picked up by scout; no pull came of it (failed) · 3 runs
4 earlier comments
Parked by the crew: the run ended without a pull: declined: this needs a person and a machine, and a crew sandbox has neither. The issue itself says so.
Parked by the crew: the run ended without a pull: declined: this needs a person and a machine, which a crew sandbox does not have. The issue asks for a run from a third machine through the production edge, against a copy restored from production's snapshot. Fr
Parked twice since a person last spoke here, so it is the planner's now. Every reason:
- the run ended without a pull: declined: this needs a person and a machine, and a crew sandbox has neither. The issue itself says so
- the run ended without a pull: declined: this needs a person and a machine, which a crew sandbox does not have. The issue asks for a run from a third machine through the production edge, against a copy restored from production's snapshot. Fr
The planner splits it, replaces it with a fresh issue, or defers it; a person can also say what to try. Nothing starts here until then, and the board lists it.
Parked by the crew: the run ended without a pull: declined: this needs a person and a machine, and a crew sandbox has neither. The issue asks for a restored copy of production's snapshot booted behind the edge, then 500/s for 60 s (after a 40 s warm-up) from a
Parked 3 times since a person last spoke here, so it is the planner's now. Every reason:
- the run ended without a pull: declined: this needs a person and a machine, and a crew sandbox has neither. The issue itself says so
- the run ended without a pull: declined: this needs a person and a machine, which a crew sandbox does not have. The issue asks for a run from a third machine through the production edge, against a copy restored from production's snapshot. Fr
- the run ended without a pull: declined: this needs a person and a machine, and a crew sandbox has neither. The issue asks for a restored copy of production's snapshot booted behind the edge, then 500/s for 60 s (after a 40 s warm-up) from a
The planner splits it, replaces it with a fresh issue, or defers it; a person can also say what to try. Nothing starts here until then, and the board lists it.
Deferred behind the Later gate (#595) by the planner. The crew tried it three times, and each run declined for the same reason: the work needs a person and a machine. Someone has to restore production's snapshot behind the edge and drive 500/s from a third machine. A crew sandbox reaches no edge, and all of its load would come from one address. The half the crew could do is done: the in-process run past the edge, kept in bench/results/9622-in-process-bypass-stale-raw-.json (pull #12099). Re-splitting it would not help: every remaining step needs the same machine. The owner can start it by unlinking it from #595 when a machine is free. The steps are in docs/features/load-test.md → The person's release run* and Still open for #9622.
Done: it is now behind the Later gate (#595). To avoid a loop, #9622 moved out from in front of P0 Trust (#586) and also went behind the gate. #12135 still holds up #9622.
The half of #9622 a crew sandbox cannot do. The in-process run past the edge is done on a copy shaped like production (83 repositories, 35,581 files, 897 MB): 500/s for 60 s, none refused, p95 38.2 ms and 20.8 ms, kept in
bench/results/9622-in-process-bypass-stale-raw-*.json(pull #12099). A sandbox reaches no edge, and through an edge its made-up client addresses are ignored, so its load would come from one address and meet the stranger gate's per-address limit. So this needs a person and a machine.Done looks like:
Never at production:
bench/load.tsrefuses it by name and by address before any request.