* refactor: extract resolve_coalesced, rewrite tests against real code
Extract Disposition enum, acquire_inflight(), and resolve_coalesced()
from handle_query so coalescing logic is independently testable. Rewrite
integration tests to call resolve_coalesced directly with mock futures
instead of fighting the iterative resolver's NS chain. All 12 coalescing
tests now exercise production code paths, not tokio primitives.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: SERVFAIL echoes question section, preserve error messages
resolve_coalesced now takes &DnsPacket instead of query_id so SERVFAIL
responses use response_from (echoing question section per RFC). Error
messages preserved via Option<String> return for upstream error logging.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* feat: in-flight query coalescing for recursive resolver
When multiple queries for the same (domain, qtype) arrive concurrently
and all miss the cache, only the first triggers recursive resolution.
Subsequent queries wait on a broadcast channel for the result.
Prevents thundering herd where N concurrent cache misses each
independently walk the full NS chain, compounding timeouts.
Uses InflightGuard (Drop impl) to guarantee map cleanup on
panic/cancellation — prevents permanent SERVFAIL poisoning.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* style: add InflightMap type alias for clippy
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: add COALESCED query path and coalescing tests
Followers in the inflight coalescing path now log as COALESCED instead
of RECURSIVE, making it visible in the dashboard when queries were
deduplicated vs independently resolved. Adds 10 tests covering
InflightGuard cleanup, broadcast mechanics, and concurrent handle_query
coalescing through a mock TCP DNS server.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* style: cargo fmt
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* refactor: extract acquire_inflight, rewrite tests against real code
Move Disposition enum and inflight acquisition logic into a standalone
acquire_inflight() function. Rewrite 4 tests that were exercising tokio
primitives to call the real coalescing code path instead.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: SRTT-based nameserver selection for recursive resolver
BIND-style Smoothed RTT (EWMA) tracking per NS IP address. The resolver
learns which nameservers respond fastest and prefers them, eliminating
cascading timeouts from slow/unreachable IPv6 servers.
- New src/srtt.rs: SrttCache with record_rtt, record_failure, sort_by_rtt
- EWMA formula: new = (old * 7 + sample) / 8, 5s failure penalty, 5min decay
- TCP penalty (+100ms) lets SRTT naturally deprioritize IPv6-over-TCP
- Enabled flag embedded in SrttCache (no-op when disabled)
- Batch eviction (64 entries) for O(1) amortized writes at capacity
- Configurable via [upstream] srtt = true/false (default: true)
- Benchmark script: scripts/benchmark.sh (full, cold, warm, compare-all)
- Benchmarks show 12x avg improvement, 0% queries >1s (was 58%)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* feat: show DNSSEC and SRTT status in dashboard + API
Add dnssec and srtt boolean fields to /stats API response.
Display on/off indicators in the dashboard footer.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: apply SRTT decay before EWMA so recovered servers rehabilitate
Without decay-before-EWMA, a server penalized at 5000ms stayed near
that value even after recovery — the stale raw penalty was used as the
EWMA base instead of the decayed estimate. Extract decayed_srtt()
helper and call it in record_rtt() before the smoothing step.
Also restores removed "why" comments in send_query / resolve_recursive.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs: add install/upgrade instructions, smarter benchmark priming
README: document `numa install`, `numa service`, Homebrew upgrade,
and `make deploy` workflows. Benchmark: replace fixed `sleep 4` with
`wait_for_priming` that polls cache entry count for stability.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
.local is reserved for mDNS (RFC 6762) and cannot be resolved by
upstream DNS servers. Add it to is_special_use_domain() so queries
like _grpc_config.localhost.local get an immediate NXDOMAIN instead
of timing out and returning SERVFAIL.
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* feat: DNS-over-HTTPS upstream forwarding
Encrypt upstream queries via DoH — ISPs see HTTPS traffic on port 443,
not plaintext DNS on port 53. URL scheme determines transport:
https:// = DoH, bare IP = plain UDP. Falls back to Quad9 DoH when
system resolver cannot be detected.
- Upstream enum (Udp/Doh) with Display and PartialEq
- BytePacketBuffer::from_bytes constructor
- reqwest http2 feature for DoH server compatibility
- network_watch_loop guards against DoH→UDP silent downgrade
- 5 new tests (mock DoH server, HTTP errors, timeout)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* style: cargo fmt
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* docs: add DoH to README — Why Numa, comparison table, roadmap
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
HTTPS proxy certs were generated once at startup. Services added at
runtime via API or LAN discovery got "not secure" in the browser
because their SAN wasn't in the cert. Now the cert is regenerated
on every service add/remove and swapped atomically via ArcSwap.
In-flight connections are unaffected.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- banner box width adapts to longest value (fixes overflow with long paths)
- hoist h() HTML escape function to script top, remove 3 local copies
- serve_ca: add Cache-Control: public, max-age=86400
- restore log path in dashboard footer alongside new config/data fields
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- SRV record uses first service's port (was 0, confused dns-sd -L)
- Remove examples/mdns_coexist.rs (served its purpose as spike)
- Reject percent-encoding in route paths (defense-in-depth)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add lan_enabled to ServerCtx
- Add lan field to /stats API (enabled, peer count)
- Dashboard shows "LAN off" (dim) or "LAN on · N peers" (green)
- Tooltip shows enable command or mDNS service type
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- numa lan on/off toggles LAN discovery in numa.toml
- Writes [lan] section if missing, updates enabled if present
- Colored output with restart hint
- README: add lan on/off to help text
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add inline route management (+ route / x) per service in dashboard
- Expose service source (config vs api) in API response
- Only show service delete button for API-created services
- Pre-fill route port with service target_port
- Fix XSS in route path onclick handlers
- Skip renderServices refresh while route form is open (editingRoute guard)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- LAN discovery disabled by default (opt-in via [lan] enabled = true)
- Replace custom JSON multicast (239.255.70.78:5390) with standard mDNS
(_numa._tcp.local on 224.0.0.251:5353) using existing DNS parser
- Instance ID in TXT record for multi-instance self-filtering
- API and proxy bind to 127.0.0.1 by default (0.0.0.0 when LAN enabled)
- Path-based routing: longest prefix match with optional prefix stripping
via [[services]] routes = [{path, port, strip?}]
- REST API: GET/POST/DELETE /services/{name}/routes
- Dashboard shows route lines per service when configured
- Segment-boundary route matching (prevents /api matching /apiary)
- Route path validation (rejects path traversal)
Closes#11
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
LAN IP checked every 5s (cheap UDP socket call). Full upstream
re-detection runs every 30s as safety net, or immediately when
LAN IP changes. Reduces worst-case network switch recovery from
30s to 5s.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Expose current upstream address in /stats response. Dashboard footer
now shows "Upstream: x.x.x.x:53" — updates live when the network
watcher swaps the upstream.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When numa install is active, scutil --dns only returns 127.0.0.1.
Previously fell back to 9.9.9.9 (Quad9) which fails on networks
that block external DNS. Now reads DHCP-provided DNS from
ipconfig getpacket en0/en1 as intermediate fallback before Quad9.
Tested on a network that blocks 8.8.8.8, 9.9.9.9, 1.1.1.1 but
allows ISP DNS (213.154.124.25) — Numa now auto-detects and uses it.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Always detect network changes (LAN IP, upstream, peers) regardless
of upstream config. LAN IP is now tracked in ServerCtx and updated
every 30s — multicast announcements use the current IP instead of
the startup IP. Upstream re-detection still only runs when
auto-detected. Peer flush triggers on any network change.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Upstream DNS was resolved once at startup and never updated. Switching
Wi-Fi networks made all queries fail until restart.
Now spawns a background task (every 30s) that re-runs system DNS
discovery and swaps the upstream atomically if it changed. Also flushes
stale LAN peers from the old network on change.
Only activates when upstream is auto-detected (not explicitly configured).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
On Windows, the not(macos/linux) cfg block is the only path, so
clippy flags the return as needless. Use expression form instead.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Gate version detection behind cfg(unix), fix unreachable Ok(()) after
return in trust_ca, use next_back() and is_some_and() per clippy.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Cross-platform paths: config_dir() uses %APPDATA%, data_dir() uses
%PROGRAMDATA% on Windows. TLS cert directory uses data_dir() instead
of hardcoded /usr/local/var/numa. Windows DNS discovery via ipconfig.
Fixed cfg gates from not(macos) to explicit linux to prevent Linux
code compiling on Windows. Added Windows target to CI and release
workflows with zip packaging.
System integration (numa install/service) not yet supported on Windows
— users run numa.exe manually.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Show whether each service is reachable from the network or bound to
localhost only. Dashboard displays green "LAN" or amber "local only"
badge next to each healthy service. Unified TCP check function,
concurrent health+LAN probes.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace IP-based self-announcement filtering with a per-process instance
ID (pid ^ timestamp) so multiple instances on the same host can discover
each other. Enable SO_REUSEPORT for multicast socket binding on Unix.
Add multicast address validation on configured group.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Numa instances on the same network auto-discover each other's .numa
services. No config, no cloud — just multicast on 239.255.70.78:5390.
- PeerStore with lazy expiry (90s timeout, 30s broadcast interval)
- DNS resolves remote .numa services to peer's LAN IP (not localhost)
- Proxy forwards to peer IP for remote services
- Graceful degradation if multicast bind fails
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Roman Stone themed 404 with Instrument Serif heading, JetBrains Mono
domain badge, brick pattern background, syntax-highlighted curl
example, and a delayed easter egg. Also updates dashboard link in
README to numa.numa.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
recv_from can never return more bytes than the buffer size — the kernel
truncates silently. == is the correct heuristic for detecting truncation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Set TC (truncation) bit when response exceeds 4096-byte buffer
instead of dropping the response silently. Clients can retry via TCP.
- Log when upstream response is truncated in forward.rs.
- Dockerfile: bump to Rust 1.88, include site/service files, use
alpine runtime instead of scratch, add cmake/perl for aws-lc-sys.
- Makefile deploy: platform-aware — codesign on macOS, systemctl on Linux.
- README: trim roadmap to near-term items only.
- Verified: Docker build + smoke test passes on Linux (Alpine musl).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
security delete-certificate -c fails when multiple certs match.
Now finds all certs by hash and deletes each individually.
Also updated README with HTTPS, service persistence, and TLS mentions.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
macOS kills unsigned binaries, so numa service restart failed after
copying a new build. Added ad-hoc codesign to restart flow and a
make deploy target that handles the full build-copy-sign-restart cycle.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
HTTP reverse proxy on port 80 lets developers use clean domain names
(frontend.numa, api.numa) instead of localhost:PORT. Includes WebSocket
upgrade support for HMR, TCP health checks, dashboard UI panel, and
REST API for service management. numa.numa is preconfigured for the
dashboard itself.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Kills the running process and lets launchd/systemd respawn it
with the updated binary. DNS stays configured throughout.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Default upstream auto-detected from system resolver (scutil/resolv.conf)
instead of hardcoding Google 8.8.8.8. Falls back to Quad9 (9.9.9.9).
- Single scutil --dns pass for both upstream detection and forwarding rules
- Linux: reads backup resolv.conf if current only has loopback
- Service start/stop now couples DNS config (install on start, uninstall on stop)
- Install script for one-line binary install from GitHub Releases
- GitHub Actions release workflow: builds for macOS/Linux x86_64/aarch64
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Root cause: upstream resolvers return EDNS OPT records (type 41) in
the additional section. Our parser reads them as UNKNOWN, but write()
silently skips them — creating a header that claims N additional records
but a body with 0, producing FORMERR on the client side.
Fix: filter out UNKNOWN records before serialization and adjust header
counts to match. Also increase BytePacketBuffer from 512 to 4096 bytes
to handle modern DNS responses with many records.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
GET /blocking/check/{domain} — returns whether a domain is blocked,
the reason (exact match, parent domain, allowlist, disabled), and
the matching rule. Dashboard sidebar has a "Check Domain" search
box with inline results and one-click allow button.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- numa service start/stop/status commands (launchd + systemd)
- Dashboard footer shows log paths (macOS + Linux) and GitHub link
- Help text updated with all service commands
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Detect systemd-resolved: use drop-in config instead of overwriting
/etc/resolv.conf (which gets regenerated)
- Warn if /etc/resolv.conf is a symlink (NetworkManager, etc.)
- Fix TOCTOU: attempt copy/remove directly, handle NotFound
- Remove side-effect from backup_path_linux (no eager mkdir)
- Fix macOS $HOME fallback: /var/root instead of /tmp
- Log warnings on launchctl/systemctl failures instead of silencing
- Delete plist before unloading (prevents zombie restarts)
- Extract ensure_binary_installed helper on Linux
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Linux:
- numa install: backs up /etc/resolv.conf, sets nameserver to 127.0.0.1
- numa uninstall: restores original /etc/resolv.conf from backup
- numa service start: installs systemd unit, enables + starts
- numa service stop: stops, disables, removes unit file
- numa service status: shows systemctl status
macOS: launchd plist (already working)
Both platforms: Restart=always / KeepAlive=true for crash recovery.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move HashMap, PathBuf, numa_data_dir, backup_path inside macOS
cfg blocks so Linux builds don't see unused imports/functions.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
