orca-control 0.2.9

Control plane: API server, reconciler, and cluster state management
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
//! Cluster networks dashboard: per-node `orca-*` Docker bridge listing plus
//! the public-edge domain routes registered with each node's proxy, and a
//! missing-alias pass that flags env references with no DNS path.

use std::collections::{HashMap, HashSet};
use std::sync::Arc;
use std::time::Duration;

use axum::Json;
use axum::extract::State;
use axum::response::IntoResponse;
use tokio::net::lookup_host;
use tracing::warn;

use orca_core::api_types::{
    ClusterNetworksResponse, DockerNetwork, DomainRoute, NodeNetworks, NodeRole,
};
use orca_core::ws_types::{MasterMessage, NetworkStatusReportData};

use crate::state::AppState;

/// Per-agent collection timeout. Generous because docker `inspect_container`
/// over many containers can be slow on a loaded box.
const NETWORK_REPORT_TIMEOUT: Duration = Duration::from_secs(5);

/// Per-domain DNS resolution timeout. Short — most resolvers answer in ms;
/// any domain that doesn't respond quickly is "unknown" for this snapshot
/// and the operator will see a `?` in the IP column.
const DNS_TIMEOUT: Duration = Duration::from_millis(800);

/// GET /api/v1/cluster/networks — aggregate Docker network + edge route info
/// across every node in the cluster, with DNS-resolved domain IPs and a
/// missing-alias pass that flags broken env references.
pub(crate) async fn cluster_networks(State(state): State<Arc<AppState>>) -> impl IntoResponse {
    // Step 1: slice the master's route table by placement node so each
    // node row gets the edge routes for services it actually hosts.
    let routes_by_node = group_routes_by_placement(&state).await;

    let master = collect_master(&state, &routes_by_node).await;
    let agents = collect_agents(&state, &routes_by_node).await;

    let mut nodes = vec![master];
    nodes.extend(agents);

    // Step 2: resolve every unique domain in parallel. Cheap-ish (DNS is
    // milliseconds for cached entries, capped by DNS_TIMEOUT otherwise).
    resolve_domain_ips(&mut nodes).await;

    // Step 3: scan service env across the cluster and flag references with
    // no matching alias on a shared network.
    annotate_missing_aliases(&mut nodes, &state).await;

    Json(ClusterNetworksResponse { nodes })
}

/// Build `placement.node → [DomainRoute]`. The key is `Some(hostname)` for
/// agent-pinned services and `None` for services with no placement (i.e.,
/// the master). Routes whose service isn't in the services map (e.g. stale
/// route after a service was removed) fall under `None` defensively.
async fn group_routes_by_placement(state: &AppState) -> HashMap<Option<String>, Vec<DomainRoute>> {
    // Snapshot the data we need under tight, scoped reads, then drop both
    // guards before any further work. Holding `state.route_table.read()`
    // across iteration was the primary lock-contention source that stalled
    // proxy TLS handshakes by up to 20 seconds — tokio::sync::RwLock uses
    // write-priority, so once *any* writer (update_container_routes from
    // the reconciler/watchdog) queues behind this long-running reader, all
    // subsequent readers (including the serve-loop check that happens
    // between peek_sni and acceptor.accept) queue behind the writer too.
    // See project_v0_2_9_rc2_proxy_lock_contention for the diagnosis.
    let service_placements: HashMap<String, Option<String>> = {
        let services = state.services.read().await;
        services
            .values()
            .map(|s| {
                (
                    s.config.name.clone(),
                    s.config.placement.as_ref().and_then(|p| p.node.clone()),
                )
            })
            .collect()
    };
    let route_snapshot: Vec<(String, Vec<String>)> = {
        let routes = state.route_table.read().await;
        routes
            .iter()
            .map(|(domain, targets)| {
                (
                    domain.clone(),
                    targets.iter().map(|t| t.service_name.clone()).collect(),
                )
            })
            .collect()
    };

    let mut out: HashMap<Option<String>, Vec<DomainRoute>> = HashMap::new();
    for (domain, service_names) in &route_snapshot {
        for service_name in service_names {
            let node = service_placements.get(service_name).cloned().flatten();
            out.entry(node).or_default().push(DomainRoute {
                domain: domain.clone(),
                service: service_name.clone(),
                resolved_ip: None,
            });
        }
    }
    for v in out.values_mut() {
        v.sort_by(|a, b| a.domain.cmp(&b.domain).then(a.service.cmp(&b.service)));
        v.dedup_by(|a, b| a.domain == b.domain && a.service == b.service);
    }
    out
}

async fn collect_master(
    _state: &AppState,
    routes_by_node: &HashMap<Option<String>, Vec<DomainRoute>>,
) -> NodeNetworks {
    let networks = enumerate_local_orca_networks().await;
    let domains = routes_by_node.get(&None).cloned().unwrap_or_default();
    NodeNetworks {
        node_id: None,
        hostname: master_hostname(),
        role: NodeRole::Master,
        networks,
        domains,
        reachable: true,
    }
}

async fn collect_agents(
    state: &AppState,
    routes_by_node: &HashMap<Option<String>, Vec<DomainRoute>>,
) -> Vec<NodeNetworks> {
    let agent_ids: Vec<u64> = state.ws_agents.read().await.keys().copied().collect();
    if agent_ids.is_empty() {
        return Vec::new();
    }

    let (tx, mut rx) = tokio::sync::mpsc::channel::<NetworkStatusReportData>(agent_ids.len() + 1);
    let mut request_ids: Vec<String> = Vec::with_capacity(agent_ids.len());
    {
        let mut listeners = state.network_listeners.write().await;
        for _ in 0..agent_ids.len() {
            let req = uuid::Uuid::new_v4().to_string();
            listeners.insert(req.clone(), tx.clone());
            request_ids.push(req);
        }
    }

    let mut dispatched = HashMap::<u64, String>::new();
    {
        let agents = state.ws_agents.read().await;
        for (node_id, req_id) in agent_ids.iter().zip(request_ids.iter()) {
            if let Some(agent_tx) = agents.get(node_id) {
                let sent = agent_tx
                    .send(MasterMessage::NetworkStatusRequest {
                        request_id: req_id.clone(),
                    })
                    .await
                    .is_ok();
                if sent {
                    dispatched.insert(*node_id, req_id.clone());
                }
            }
        }
    }

    let mut reports = HashMap::<u64, NetworkStatusReportData>::new();
    let deadline = tokio::time::sleep(NETWORK_REPORT_TIMEOUT);
    tokio::pin!(deadline);
    while reports.len() < dispatched.len() {
        tokio::select! {
            biased;
            msg = rx.recv() => {
                match msg {
                    Some(data) => { reports.insert(data.node_id, data); }
                    None => break,
                }
            }
            _ = &mut deadline => {
                warn!(
                    "cluster_networks: timed out after {}s ({} of {} agents responded)",
                    NETWORK_REPORT_TIMEOUT.as_secs(),
                    reports.len(),
                    dispatched.len(),
                );
                break;
            }
        }
    }

    {
        let mut listeners = state.network_listeners.write().await;
        for req in &request_ids {
            listeners.remove(req);
        }
    }

    agent_ids
        .into_iter()
        .map(|node_id| match reports.remove(&node_id) {
            Some(data) => {
                let domains = routes_by_node
                    .get(&Some(data.hostname.clone()))
                    .cloned()
                    .unwrap_or_default();
                NodeNetworks {
                    node_id: Some(node_id),
                    hostname: data.hostname,
                    role: NodeRole::Agent,
                    networks: data.networks,
                    domains,
                    reachable: true,
                }
            }
            None => NodeNetworks {
                node_id: Some(node_id),
                hostname: format!("node-{node_id}"),
                role: NodeRole::Agent,
                networks: Vec::new(),
                domains: Vec::new(),
                reachable: false,
            },
        })
        .collect()
}

/// Walk every domain across every node in parallel and populate
/// `resolved_ip`. Bounded by [`DNS_TIMEOUT`] per domain so a stuck
/// resolver doesn't block the dashboard. Uses the OS resolver via
/// `tokio::net::lookup_host`.
async fn resolve_domain_ips(nodes: &mut [NodeNetworks]) {
    use futures_util::future::join_all;

    // Dedup domains so we don't pay multiple times for the same name.
    let mut unique: HashSet<String> = HashSet::new();
    for n in nodes.iter() {
        for d in &n.domains {
            unique.insert(d.domain.clone());
        }
    }
    let jobs = unique.into_iter().map(|domain| async move {
        let target = format!("{domain}:0");
        let result = tokio::time::timeout(DNS_TIMEOUT, lookup_host(target)).await;
        let ip = match result {
            Ok(Ok(mut iter)) => iter.next().map(|sa| sa.ip().to_string()),
            _ => None,
        };
        (domain, ip)
    });
    let resolved: HashMap<String, Option<String>> = join_all(jobs).await.into_iter().collect();
    for n in nodes.iter_mut() {
        for d in &mut n.domains {
            if let Some(ip) = resolved.get(&d.domain).and_then(|x| x.clone()) {
                d.resolved_ip = Some(ip);
            }
        }
    }
}

/// Heuristic missing-alias detection. For each service that has env vars,
/// look for tokens matching another known service's name. If the referenced
/// service isn't aliased as that name on a shared Docker network, record the
/// missing alias on the *referenced* service's row so the dashboard can
/// highlight where the gap is.
///
/// The detection is intentionally conservative — false positives are worse
/// than false negatives for an ops dashboard (the operator stops trusting
/// the warning), so we only flag exact substring matches that look like
/// hostnames (whole words bounded by typical URL chars).
async fn annotate_missing_aliases(nodes: &mut [NodeNetworks], state: &AppState) {
    let services = state.services.read().await;
    if services.is_empty() {
        return;
    }
    // service_name → its set of aliases across all networks. Used both as
    // "is this name a known service?" and to compare references against.
    let mut all_aliases: HashMap<String, HashSet<String>> = HashMap::new();
    for n in nodes.iter() {
        for net in &n.networks {
            for svc in &net.services {
                let entry = all_aliases.entry(svc.name.clone()).or_default();
                for a in &svc.aliases {
                    entry.insert(a.clone());
                }
            }
        }
    }
    // Build a referrer map: referenced_service → set of names it's looked
    // up as. A name is "looked up as X" when service A's env contains a
    // bare token X that matches another service name.
    let mut references: HashMap<String, HashSet<String>> = HashMap::new();
    let known_names: HashSet<&str> = services.keys().map(|s| s.as_str()).collect();
    for svc in services.values() {
        for value in svc.config.env.values() {
            for token in tokens_from_env_value(value) {
                if known_names.contains(token.as_str()) && token != svc.config.name {
                    references.entry(token.clone()).or_default().insert(token);
                }
            }
        }
    }
    // For each container row in the rendered networks, see whether its
    // aliases cover everything it's referenced as. The container name
    // produced by orca is `orca-<service>` or `orca-<service>-<n>`; map
    // back to the service name by trimming.
    for n in nodes.iter_mut() {
        for net in &mut n.networks {
            for svc in &mut net.services {
                let svc_short = service_name_from_container(&svc.name);
                let Some(refs) = references.get(&svc_short) else {
                    continue;
                };
                let alias_set: HashSet<&str> = svc.aliases.iter().map(|s| s.as_str()).collect();
                let missing: Vec<String> = refs
                    .iter()
                    .filter(|r| !alias_set.contains(r.as_str()))
                    .cloned()
                    .collect();
                if !missing.is_empty() {
                    let mut m = missing;
                    m.sort();
                    m.dedup();
                    svc.missing_aliases = m;
                }
            }
        }
    }
}

/// Extract bare-word tokens from an env value that *look* like hostnames.
/// Splits on common separators in URLs and connection strings; keeps tokens
/// composed of `[a-z0-9-]` characters. Aggressive enough to catch
/// `postgres://db:5432/x` (extracts "db") and `http://api/v1` (extracts
/// "api"), but conservative enough not to fire on prose values.
pub(crate) fn tokens_from_env_value(value: &str) -> Vec<String> {
    let mut out: Vec<String> = Vec::new();
    let mut cur = String::new();
    for ch in value.chars() {
        if ch.is_ascii_lowercase() || ch.is_ascii_digit() || ch == '-' {
            cur.push(ch);
        } else {
            if !cur.is_empty() && cur.len() > 1 {
                out.push(std::mem::take(&mut cur));
            }
            cur.clear();
        }
    }
    if !cur.is_empty() && cur.len() > 1 {
        out.push(cur);
    }
    out
}

/// Orca container names are `orca-<service>` (and `orca-<service>-<N>` for
/// replicas). Strip the prefix and trailing replica suffix so we can match
/// against `services.keys()`.
fn service_name_from_container(container: &str) -> String {
    let rest = container.trim_start_matches("orca-");
    // Drop a trailing `-<digits>` suffix if present.
    if let Some((stem, tail)) = rest.rsplit_once('-')
        && tail.chars().all(|c| c.is_ascii_digit())
    {
        return stem.to_string();
    }
    rest.to_string()
}

/// Walk the master's own Docker daemon. Delegates to the agent crate's
/// enumerator so master and agent always see the same shape.
async fn enumerate_local_orca_networks() -> Vec<DockerNetwork> {
    orca_agent::ws_client::network_status::list_local_orca_networks().await
}

fn master_hostname() -> String {
    std::fs::read_to_string("/etc/hostname")
        .map(|s| s.trim().to_string())
        .or_else(|_| std::env::var("HOSTNAME"))
        .unwrap_or_else(|_| "master".to_string())
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn tokens_extract_hostnames_from_url() {
        assert_eq!(
            tokens_from_env_value("postgres://db:5432/app"),
            vec!["postgres", "db", "5432", "app"]
        );
        assert_eq!(
            tokens_from_env_value("http://api-gateway/v1"),
            vec!["http", "api-gateway", "v1"]
        );
    }

    #[test]
    fn tokens_skip_single_chars_and_uppercase_garbage() {
        // Single-char tokens are dropped (`len > 1`), and uppercase
        // content (`A`, `B`) doesn't contribute since we match `[a-z0-9-]`
        // only. So `c` is excluded too — single char.
        assert_eq!(tokens_from_env_value("A=B;c=d-e"), vec!["d-e"]);
    }

    #[test]
    fn service_name_strips_orca_prefix_and_replica_suffix() {
        assert_eq!(service_name_from_container("orca-api"), "api");
        assert_eq!(service_name_from_container("orca-api-3"), "api");
        assert_eq!(
            service_name_from_container("orca-api-gateway-1"),
            "api-gateway"
        );
        // No prefix → pass through. Trailing non-numeric suffix → keep the
        // service-name part after stripping `orca-`.
        assert_eq!(service_name_from_container("standalone"), "standalone");
        assert_eq!(service_name_from_container("orca-api-canary"), "api-canary");
    }
}