dial9-viewer

Library crate backing the dial9 CLI. Install and use via the dial9 crate:

cargo install --locked dial9

See the dial9 README for usage documentation.

Simulator mode

Run the complete S3-browser and on-demand aggregation workflow without an S3 bucket or AWS credentials:

# Sanitized synthetic traces
dial9 serve --simulator --local

# Rebase and replay the bundled demo trace in every virtual segment
dial9 serve --simulator demo --local

# Configure fleet size, segment spacing, data volume, and feature groups
dial9 serve --simulator synthetic \
  --simulator-hosts 8 \
  --simulator-segment-secs 60 \
  --simulator-repetitions 2 \
  --simulator-symbols realistic \
  --simulator-features cpu,scheduling,tasks,spans \
  --local

Simulator objects use the production date/service/host key layout and the normal viewer storage API. The catalog is virtual: every requested time range has deterministic segments, and trace bytes are rendered only when fetched. Flamegraph, span, and Tokio-stat rollups are written to a process-local temporary directory and removed when the server exits. Run dial9 serve --help for all simulator size, duration, and symbol options. Synthetic symbol names remain anonymous placeholders by default; --simulator-symbols realistic emits deterministic Rust-like names for more representative flamegraphs.

trace-shape — Trace Structural Fingerprints

The trace-shape subcommand extracts and generates sanitized structural fingerprints ("shapes") from dial9 traces. Shapes preserve the operational characteristics of a trace—event types, timing distributions, cardinality, field schemas—with best-effort removal of labels, identifiers, payloads, and exact timestamps. This is not an anonymization or security boundary.

Usage

# Sanitize directly into a synthetic trace (preferred for large traces)
dial9 trace-shape synthesize /tmp/traces/trace.bin synthetic.bin

# Repeat the in-memory template 5 times without writing shape JSON
dial9 trace-shape synthesize /tmp/traces/trace.bin synthetic.bin --repeat 5

# Extract a portable shape from a trace file (accepts gzip input)
dial9 trace-shape extract /tmp/traces/trace.bin shape.json

# Generate a synthetic trace from a previously extracted shape
dial9 trace-shape generate shape.json synthetic.bin

trace-shape synthesize builds the same sanitized shape in memory and writes its synthetic trace directly. It avoids the verbose per-event JSON intermediate and its input-size limit, while preserving the same privacy transformations, correlations, validation, and repeat behavior as extract followed by generate.

What shapes preserve

• Event types, field schemas, and field types
• Relative timing and event ordering (quantized to 10 µs)
• Worker and task cardinality
• Value magnitudes and distributions (quantized)
• Byte payload lengths, stack depths, and dynamic-container cardinalities
• String values replaced with fixed deterministic placeholders (source lengths are not retained)
• Correlated identity namespaces (task, span, thread, address)
• Built-in schema/field names needed by the viewer

What shapes remove

• Source absolute timestamps (monotonic and realtime)
• String contents (replaced with fixed-length deterministic placeholders, e.g. s_0001)
• Custom schema and field names (replaced with anonymous names)
• Stack addresses (replaced with synthetic deterministic addresses)
• Byte payloads (zeroed, length preserved)
• Arbitrary annotation text (only allowlisted unit annotations — ns, us, ms, s, bytes, count — are retained)

Fixed-width normalization

The trace format's U8, U16, and U32 wire types are normalized to varints in shape files. The Encoder dispatches on FieldValue variants (not schema FieldType), so decoded Varint values cannot be re-encoded as fixed-width. Hand-authored shapes using fixed-width tags are rejected at validation time.

Privacy caveat

Shape extraction applies deterministic transformations to remove string contents, byte payloads, exact timestamps, custom names, and stack addresses. However, exact booleans, small quantized integers in built-in schemas, and already-round floats may survive transformation. This is not an anonymization or security boundary. Shapes intentionally retain sensitive operational structure including relative timing, event ordering, cardinality, byte payload sizes, stack depths, value magnitude distributions, and inter-event correlations. This operational structure may reveal performance characteristics, traffic patterns, or architectural details. Synthetic traces generated from shapes should be treated as confidential operational data and shared only with the same caution as production metrics.