Zachery Aaron Shores-Chmielewski
ae9ca3bcf3
Promote pipeline-parallel-inference to a first-class app and consolidate observability on the datastream wire, decoupling the dashboard crate from `distribution`. - apps/pipeline-parallel-inference: move the example out of `examples/` into `apps/` as its own workspace, rename binaries to `pp-worker`/`pp-orchestrator`, and strip release binaries - cluster: add `ClusterNode`, a synchronous facade over the actorized distribution protocol (IrohDriver + per-node Runtime hosting Swim/Registry/Metadata/Directory actors with a `MembershipFanout`), replacing ad-hoc `driver.node()`/`tick()` call sites - fleet: add per-node fleet telemetry that ships identity/resource records as `DatastreamFrame`s over the cluster transport to the orchestrator's `DatastreamSink`, folded into a `FleetView` on a 3s tick - provision: add best-effort, opt-in SSH boot-phase telemetry (`PP_DEPLOY_KEY`) that streams rented-node boot logs onto the orchestrator's datastream as `proc.boot.<stage>.*` - dashboard: rewire the crate dependency from `distribution` to `datastream`, drop the standalone `swactor-datastream-dashboard` binary, and rewrite `datastream_source.rs` to demux per-node frames into Overview/Distribution/Fleet views with live-node TTL filtering - distribution: refresh dist/netmap plugin copy and README from "Kademlia routing" to gossip-directory terminology Signed-off-by: Zachery Aaron Shores-Chmielewski <zacheryasc@gmail.com>
3.1 KiB
dashboard
Visual dashboard for the swactor runtime. Provides a live HTTP dashboard, a terminal UI (TUI), trace recording/replay, and an HTTP API for programmatic runtime investigation.
Features
| Feature | Default | Description |
|---|---|---|
distribution |
yes | /distribution page with SWIM membership, gossip directory routes, and location cache |
tui |
no | Terminal UI with overview, worker detail, and distribution views |
HTTP Dashboard
Start the dashboard demo and open it in a browser:
cargo run -p dashboard --example dashboard_demo
Pages:
http://localhost:9090— live overview (workers, actors, message rates)http://localhost:9090/actors— actor tablehttp://localhost:9090/distribution— SWIM membership, gossip directory routes, cache entries
The demo creates a 4-worker runtime with ping-pong and counter actors, plus a 9-node distribution cluster (1 main node + 8 peers) with simulated SWIM membership and actor registrations in the directory/cache.
TUI
A standalone binary that connects to any running dashboard over SSE:
cargo run -p dashboard --features tui --bin swactor-tui
# or point at a specific endpoint
cargo run -p dashboard --features tui --bin swactor-tui -- http://localhost:9090
Views (cycle with Tab):
- Overview — htop-style worker bars, summary line, sortable actor table
- Worker Detail — focused view of a single worker's actors and phase breakdown
- Distribution — cluster summary, scrollable members table, cache entries, gossip directory route count
Key bindings: q quit, Tab cycle views, s sort column, r reverse sort,
arrow keys/j/k scroll, Enter drill into worker, Esc back to overview.
Agent HTTP API (Investigate)
All diagnostic commands are available as HTTP endpoints when the dashboard server is running. See AGENTS.md for full protocol documentation.
curl 'http://localhost:9090/api/investigate?cmd=overview'
curl 'http://localhost:9090/api/investigate?cmd=hot&n=5'
curl 'http://localhost:9090/api/investigate?cmd=workers'
curl 'http://localhost:9090/api/investigate?cmd=worker&id=2'
curl 'http://localhost:9090/api/investigate?cmd=actors&sort=mailbox&limit=10'
curl 'http://localhost:9090/api/investigate?cmd=diff&seconds=2'
The same commands are also available via a stdin/stdout REPL for direct
programmatic use (see investigate::run_investigate).
Demos
All examples are run from the workspace root.
HTTP dashboard — live workload with distribution cluster, Ctrl+C to stop:
cargo run -p dashboard --example dashboard_demo
# http://localhost:9090 — runtime overview
# http://localhost:9090/distribution — cluster view
Benchmarks — four automated scenarios (~20 s total):
cargo run -p dashboard --example bench_dashboard
# open http://localhost:9090
Record & replay — records ~10 s of activity, then serves a replay:
cargo run -p dashboard --example record_and_replay_demo
# live dashboard at http://localhost:9090 during recording
# replay dashboard at http://localhost:9091 after recording finishes
# Ctrl+C to stop