fix(perf): accept summaries up to 192 KiB #17

Merged
claude merged 1 commit from codex/perf-body-192k into main 2026-11-17 02:58:14 -05:00
Collaborator

With one sample a second (Longshot#1372), a performance summary is about 172 bytes per sample. At 64 KiB the relay refused anything over about 370 samples:

  • the default 5-minute summary fits with little room: 54.5 KB with 4 maps, 62.2 KB with 16 (the most a summary holds);
  • any RX_VK_PERF_INTERVAL over about 6 minutes is refused with 413 and its data lost (600 s came to 106 KB), although MaxSamples is 900 and the game allows up to 14 minutes.

Measured with the game's own summary_json() against a local relay.

Changes

  • telemetry/perf: MaxBody 64 KiB → 192 KiB, which fits 900 samples across 16 maps (about 165 KB).
  • The generated OpenAPI and reference docs, and the deployment diagram, follow.

Bean-Ops' perf-body-limit Traefik middleware is raised to match at the same time. Raising it before this relay is deployed is harmless: the relay keeps answering 413 over 64 KiB until then.

Checks run locally

  • go test -race -count=1 ./...: pass
  • go test -tags e2e -count=1 ./test/e2e/...: pass
  • scripts/lint.sh: 0 issues
  • scripts/docs.sh --check: current

🤖 Generated with Claude Code

https://claude.ai/code/session_015KgSKmkvJhEPjxuHpnkA9s

With one sample a second (Longshot#1372), a performance summary is about 172 bytes per sample. At 64 KiB the relay refused anything over about 370 samples: - the default 5-minute summary fits with little room: 54.5 KB with 4 maps, 62.2 KB with 16 (the most a summary holds); - any `RX_VK_PERF_INTERVAL` over about 6 minutes is refused with 413 and its data lost (600 s came to 106 KB), although `MaxSamples` is 900 and the game allows up to 14 minutes. Measured with the game's own `summary_json()` against a local relay. ## Changes - `telemetry/perf`: `MaxBody` 64 KiB → 192 KiB, which fits 900 samples across 16 maps (about 165 KB). - The generated OpenAPI and reference docs, and the deployment diagram, follow. Bean-Ops' `perf-body-limit` Traefik middleware is raised to match at the same time. Raising it before this relay is deployed is harmless: the relay keeps answering 413 over 64 KiB until then. ## Checks run locally - `go test -race -count=1 ./...`: pass - `go test -tags e2e -count=1 ./test/e2e/...`: pass - `scripts/lint.sh`: 0 issues - `scripts/docs.sh --check`: current 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_015KgSKmkvJhEPjxuHpnkA9s
fix(perf): accept summaries up to 192 KiB
All checks were successful
ci / lint (pull_request) Successful in 43s
ci / test (pull_request) Successful in 40s
ci / e2e (pull_request) Successful in 46s
ci / docs (pull_request) Successful in 44s
ci / image (pull_request) Successful in 31s
5c341b2d53
At 64 KiB a summary held about 370 one-second samples: the default 5-minute
interval fit with little room (62 KB at 16 maps) and any interval past about
6 minutes was refused with 413, although MaxSamples is 900 and the game allows
up to 14 minutes. 192 KiB fits 900 samples across 16 maps (about 165 KB).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015KgSKmkvJhEPjxuHpnkA9s
claude merged commit eadbca4c0b into main 2026-11-17 02:58:14 -05:00
Sign in to join this conversation.
No reviewers
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
MagicBean-Mirror/Longshot-api!17
No description provided.