JuliusBrussee/caveman · error

cachebench: observation request %q body digest mismatch

Error message

cachebench: observation request %q body digest mismatch

What it means

The final join check: an observation's RequestBodySHA256 must be non-empty and equal to the trace record's BodySHA256. A mismatch means the body the provider actually saw differs from the body the trace recorded, invalidating any cache-reuse attribution.

Source

Thrown at cacheengine/cachebench/observed.go:65

		}
		if strings.TrimSpace(request.Provider) == "" || strings.TrimSpace(request.Epoch) == "" || !validUniqueJSONObject(request.Body) || request.BodySHA256 == "" || request.BodySHA256 != bodyDigest(request.Body) {
			return Report{}, fmt.Errorf("cachebench: trace request %q has invalid identity, body, or digest", request.RequestID)
		}
		if request.StableSegmentCount < 0 || request.StableSegmentCount > len(request.Prefix) || !validPrefix(request.Prefix) {
			return Report{}, fmt.Errorf("cachebench: trace request %q has invalid prefix", request.RequestID)
		}
		byID[request.RequestID] = request
	}
	for _, record := range records {
		request, exists := byID[record.RequestID]
		if !exists {
			return Report{}, fmt.Errorf("cachebench: observation request %q absent from trace", record.RequestID)
		}
		if record.Provider != request.Provider || record.Epoch != request.Epoch {
			return Report{}, fmt.Errorf("cachebench: observation request %q provider/epoch mismatch", record.RequestID)
		}
		if record.RequestBodySHA256 == "" || record.RequestBodySHA256 != request.BodySHA256 {
			return Report{}, fmt.Errorf("cachebench: observation request %q body digest mismatch", record.RequestID)
		}
	}
	report, err := EvaluateObserved(records, target)
	if err != nil {
		return Report{}, err
	}
	report.EvidenceLimitations = append([]string{
		"observation population exactly joined to supplied request trace by request ID and body SHA-256",
	}, report.EvidenceLimitations...)
	return report, nil
}

// ObservationReadLimits bounds JSONL evidence decoding.
type ObservationReadLimits struct {
	MaxLineBytes int
	MaxRecords   int
}

View on GitHub (pinned to 27d5a3981a)

Solutions

  1. Replay with byte-identical bodies: pass the trace's raw body bytes through untouched.
  2. Recompute the trace digest from the exact bytes the replay sends, or regenerate both together.
  3. Remove intermediaries that normalize/reformat JSON between trace capture and observation.
Defensive patterns

Strategy: validation

Validate before calling

func digestsAlign(records []ObservationRecord, trace []TraceRecord) error {
	byID := map[string]TraceRecord{}
	for _, t := range trace {
		byID[t.RequestID] = t
	}
	for _, r := range records {
		if t, ok := byID[r.RequestID]; ok && r.RequestBodySHA256 != t.BodySHA256 {
			return fmt.Errorf("observation %q body digest mismatch", r.RequestID)
		}
	}
	return nil
}

Prevention

When it happens

Trigger: EvaluateObservedAgainstTrace where record.RequestBodySHA256 is empty or differs from request.BodySHA256 — e.g. the replay sent a re-serialized body (key order or whitespace changed) after the trace digest was computed.

Common situations: Body re-encoding between trace generation and replay (map iteration order in Go, JSON pretty-printers, proxy reformatting); observations captured against a different corpus revision than the trace.

Related errors


AI-assisted analysis of JuliusBrussee/caveman@27d5a3981a (2026-08-15). Data as JSON: /api/errors/df46deb2df3efab0. Report an issue: GitHub.