fleet/fleetd
3
0
Fork 0
Code Issues 73 Pull Requests Actions Packages Projects Releases 1 Wiki Activity

fleetd #562: surface loop health
CI / shell-tests (pull_request) Successful in 12s
Details
CI / contract (pull_request) Successful in 1m31s
Details
CI / build (pull_request) Successful in 2m11s
Details

Browse Source
This commit is contained in:
Dai Ha
2026-09-12 19:26:37 +07:00
parent 204da67d66
commit 275ac0d251
6 changed files with 179 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CLAUDE.md
+1 -1
View File
@@ -144,7 +144,7 @@ the merge — and merging on a reviewer's word is delegating it by proxy.
| Confirm your own role | `fleet_whoami` |
| See backends available | `fleet_profiles` |
| Start a member | `fleet_spawn{role?, profile?, cwd?, worktree?, ticket?, sessionName?, resumeSessionId?}` → `sessionId` + `paneId` |
| See the fleet | `fleet_list` → `leads` (your peers) + `members` (each carries `agentSessionId` when its backend knows one) · one peer's state: `fleet_status{sessionId}` |
| See the fleet | `fleet_list` → `leads` (your peers) + `members` (each carries `agentSessionId` when its backend knows one) + `loopHealth` (`RUNNING`, `STALLED`, or `STOPPED` for `statusPoller` and `sessionReaper`) · one peer's state: `fleet_status{sessionId}` |
| Delegate (blocking) | `fleet_send{sessionId, content}` |
| Delegate (long task) | `fleet_send{sessionId, content, wait:false}` → ticket → `fleet_poll{ticket}` |
| Answer a member's `fleet_ask` | `fleet_send{turnId, content}` — **not** `sessionId` |
fleetd/src/main/java/dev/ltms/fleet/Fleetd.java
+5 -1
View File
@@ -21,6 +21,7 @@ import dev.ltms.fleet.inject.ExhaustedPatternLookup;
import dev.ltms.fleet.inject.ExhaustionSink;
import dev.ltms.fleet.inject.LiveExhaustedPatterns;
import dev.ltms.fleet.inject.Injector;
import dev.ltms.fleet.inject.LoopWatchdog;
import dev.ltms.fleet.inject.StatusPoller;
import dev.ltms.fleet.inject.TurnListener;
import dev.ltms.fleet.inject.MemberPresence;
@@ -665,10 +666,13 @@ public final class Fleetd {
return configured == null ? null : configured.effectiveCredentialId();
}, outagePolicy);
FleetMcp.LoopHealthSource loopHealth = new FleetMcp.LoopHealthSource(poller::health,
() -> reaper == null ? LoopWatchdog.State.STOPPED : reaper.health());
FleetMcp mcp = new FleetMcp(messages, workers, sessions, identity, presence,
primaryRegistry, callers, FleetMcp.AuthorizationMode.ENFORCED, metrics,
capacitySource(config, cfg, profile -> liveCountRef.get().apply(profile)),
healthCoverageSource(config),
loopHealth,
quarantineSource,
leadMailbox,
outageSource,
@@ -760,7 +764,7 @@ public final class Fleetd {
Javalin app = new FleetApp(herdr, memberHerdr, workers, sessions, messages, presence, mcp.servlet(),
callers, metrics, deliverable,
() -> MemberCredentialPolicyView.of(config.get().memberCredentials()),
quarantineSource, outageSource).build();
quarantineSource, outageSource, loopHealth).build();
app.start(cfg.bind().host(), cfg.bind().port());
log.info("fleetd listening on {}:{}, herdr socket {}",
cfg.bind().host(), cfg.bind().port(), socket);
fleetd/src/main/java/dev/ltms/fleet/mcp/FleetMcp.java
+76 -22
View File
@@ -20,6 +20,7 @@ import dev.ltms.fleet.msg.Rendezvous;
import dev.ltms.fleet.peer.PeerUnreachableException;
import dev.ltms.fleet.placement.BackendOutagePolicy;
import dev.ltms.fleet.placement.BackendQuarantine;
import dev.ltms.fleet.inject.LoopWatchdog;
import dev.ltms.fleet.placement.PlacementException;
import dev.ltms.fleet.session.SessionManager;
import dev.ltms.fleet.session.MemberSession;
@@ -108,6 +109,7 @@ public final class FleetMcp {
private final Metrics metrics; // CB-502: null → auth failures not counted
private final CapacitySource capacity;
private final HealthCoverageSource healthCoverage;
private final LoopHealthSource loopHealth;
private final QuarantineSource quarantine;
/** fleetd #201 Unit 5: SEPARATE from {@link #quarantine} — see {@link OutageSource}'s doc. */
private final OutageSource outage;
@@ -136,6 +138,15 @@ public final class FleetMcp {
/** Coverage is supplied by the health wiring, not inferred from a missing dependency. */
public record HealthCoverageSource(Supplier<String> value) { }
/** Progress states for fleetd's singleton background loops, read by {@code fleet_list} and {@code /healthz}. */
public record LoopHealthSource(Supplier<LoopWatchdog.State> statusPoller,
Supplier<LoopWatchdog.State> sessionReaper) {
/** Inert source for callers that do not wire the background loops. */
public static LoopHealthSource none() {
return new LoopHealthSource(() -> LoopWatchdog.State.STOPPED, () -> LoopWatchdog.State.STOPPED);
}
}
/**
* CB-578 stage B quarantine facts used by {@code fleet_profiles}: a profile → credential id
* lookup, plus the shared {@link BackendQuarantine} to read remaining cooldowns off.
@@ -328,11 +339,22 @@ public final class FleetMcp {
* instead of throwing. See {@link #handover}.
*/
public FleetMcp(MessageService messages, PeerLauncher workers, SessionManager sessions,
ConnectionIdentity identity, MemberPresence presence, PrimaryRegistry primaryRegistry,
CallerResolver callers, AuthorizationMode authorizationMode, Metrics metrics,
CapacitySource capacity, HealthCoverageSource healthCoverage,
QuarantineSource quarantine, LeadChannel leadChannel, OutageSource outage,
LeadSeatSource leadSeats, List<String> peers, LeadRollover leadRollover) {
ConnectionIdentity identity, MemberPresence presence, PrimaryRegistry primaryRegistry,
CallerResolver callers, AuthorizationMode authorizationMode, Metrics metrics,
CapacitySource capacity, HealthCoverageSource healthCoverage,
QuarantineSource quarantine, LeadChannel leadChannel, OutageSource outage,
LeadSeatSource leadSeats, List<String> peers, LeadRollover leadRollover) {
this(messages, workers, sessions, identity, presence, primaryRegistry, callers, authorizationMode, metrics,
capacity, healthCoverage, LoopHealthSource.none(), quarantine, leadChannel, outage, leadSeats,
peers, leadRollover);
}
public FleetMcp(MessageService messages, PeerLauncher workers, SessionManager sessions,
ConnectionIdentity identity, MemberPresence presence, PrimaryRegistry primaryRegistry,
CallerResolver callers, AuthorizationMode authorizationMode, Metrics metrics,
CapacitySource capacity, HealthCoverageSource healthCoverage, LoopHealthSource loopHealth,
QuarantineSource quarantine, LeadChannel leadChannel, OutageSource outage,
LeadSeatSource leadSeats, List<String> peers, LeadRollover leadRollover) {
Objects.requireNonNull(callers, "callers");
this.authorizationEnforced = Objects.requireNonNull(authorizationMode, "authorizationMode")
== AuthorizationMode.ENFORCED;
@@ -343,6 +365,7 @@ public final class FleetMcp {
this.outage = Objects.requireNonNull(outage, "outage");
this.leadSeats = Objects.requireNonNull(leadSeats, "leadSeats");
this.healthCoverage = healthCoverage;
this.loopHealth = Objects.requireNonNull(loopHealth, "loopHealth");
this.leadRollover = leadRollover;
McpJsonMapper json = new JacksonMcpJsonMapperSupplier().get();
this.transport = HttpServletStreamableServerTransportProvider.builder()
@@ -474,7 +497,7 @@ public final class FleetMcp {
(exchange, _) -> {
McpSchema.CallToolResult denied = deny(exchange, toolAction("fleet_list", Map.of()), null);
if (denied != null) return denied;
return listFleet(workers, sessions, messages, capacity, healthCoverage, quarantine, outage,
return listFleet(workers, sessions, messages, capacity, healthCoverage, loopHealth, quarantine, outage,
leadSeats, callers.leads(),
callerTerminal(exchange),
new CoordinationSource(leadChannel, peers),
@@ -1546,25 +1569,42 @@ public final class FleetMcp {
* @param selfTerm the calling pane's terminal id, or blank for a caller with no pane
*/
static McpSchema.CallToolResult listFleet(PeerLauncher workers, SessionManager sessions,
Map<String, String> leads, String selfTerm) {
Map<String, String> leads, String selfTerm) {
return listFleet(workers, sessions, null, CapacitySource.none(), new HealthCoverageSource(() -> "off"),
QuarantineSource.none(), leads, selfTerm);
LoopHealthSource.none(), QuarantineSource.none(), leads, selfTerm);
}
static McpSchema.CallToolResult listFleet(PeerLauncher workers, SessionManager sessions, MessageService messages,
CapacitySource capacity, HealthCoverageSource healthCoverage,
QuarantineSource quarantine, Map<String, String> leads, String selfTerm) {
return listFleet(workers, sessions, messages, capacity, healthCoverage, LoopHealthSource.none(), quarantine, leads, selfTerm,
CoordinationSource.none());
}
static McpSchema.CallToolResult listFleet(PeerLauncher workers, SessionManager sessions, MessageService messages,
CapacitySource capacity, HealthCoverageSource healthCoverage,
QuarantineSource quarantine, Map<String, String> leads, String selfTerm) {
return listFleet(workers, sessions, messages, capacity, healthCoverage, quarantine, leads, selfTerm,
LoopHealthSource loopHealth, QuarantineSource quarantine,
Map<String, String> leads, String selfTerm) {
return listFleet(workers, sessions, messages, capacity, healthCoverage, loopHealth, quarantine, leads, selfTerm,
CoordinationSource.none());
}
static McpSchema.CallToolResult listFleet(PeerLauncher workers, SessionManager sessions, MessageService messages,
CapacitySource capacity, HealthCoverageSource healthCoverage,
LoopHealthSource loopHealth, QuarantineSource quarantine,
Map<String, String> leads, String selfTerm,
CoordinationSource coordination) {
return listFleet(workers, sessions, messages, capacity, healthCoverage, loopHealth, quarantine,
OutageSource.none(), LeadSeatSource.none(), leads, selfTerm, coordination, false);
}
/** As above, plus fleetd #201 Unit 5 cool-off facts (see {@link OutageSource}). */
static McpSchema.CallToolResult listFleet(PeerLauncher workers, SessionManager sessions, MessageService messages,
CapacitySource capacity, HealthCoverageSource healthCoverage,
QuarantineSource quarantine, OutageSource outage,
Map<String, String> leads, String selfTerm) {
return listFleet(workers, sessions, messages, capacity, healthCoverage, quarantine, outage,
LeadSeatSource.none(), leads, selfTerm, CoordinationSource.none());
return listFleet(workers, sessions, messages, capacity, healthCoverage, LoopHealthSource.none(), quarantine, outage,
LeadSeatSource.none(), leads, selfTerm, CoordinationSource.none(), false);
}
/**
@@ -1580,8 +1620,8 @@ public final class FleetMcp {
CapacitySource capacity, HealthCoverageSource healthCoverage,
QuarantineSource quarantine, Map<String, String> leads, String selfTerm,
CoordinationSource coordination) {
return listFleet(workers, sessions, messages, capacity, healthCoverage, quarantine, OutageSource.none(),
LeadSeatSource.none(), leads, selfTerm, coordination);
return listFleet(workers, sessions, messages, capacity, healthCoverage, LoopHealthSource.none(), quarantine, OutageSource.none(),
LeadSeatSource.none(), leads, selfTerm, coordination, false);
}
/** As above, plus fleetd #201 Unit 5 cool-off facts (see {@link OutageSource}). */
@@ -1589,8 +1629,8 @@ public final class FleetMcp {
CapacitySource capacity, HealthCoverageSource healthCoverage,
QuarantineSource quarantine, OutageSource outage,
Map<String, String> leads, String selfTerm, CoordinationSource coordination) {
return listFleet(workers, sessions, messages, capacity, healthCoverage, quarantine, outage,
LeadSeatSource.none(), leads, selfTerm, coordination);
return listFleet(workers, sessions, messages, capacity, healthCoverage, LoopHealthSource.none(), quarantine, outage,
LeadSeatSource.none(), leads, selfTerm, coordination, false);
}
/**
@@ -1611,7 +1651,7 @@ public final class FleetMcp {
QuarantineSource quarantine, OutageSource outage,
LeadSeatSource leadSeats, Map<String, String> leads, String selfTerm,
CoordinationSource coordination) {
return listFleet(workers, sessions, messages, capacity, healthCoverage, quarantine, outage,
return listFleet(workers, sessions, messages, capacity, healthCoverage, LoopHealthSource.none(), quarantine, outage,
leadSeats, leads, selfTerm, coordination, false);
}
@@ -1631,8 +1671,18 @@ public final class FleetMcp {
* an explicit {@code true}
*/
static McpSchema.CallToolResult listFleet(PeerLauncher workers, SessionManager sessions, MessageService messages,
CapacitySource capacity, HealthCoverageSource healthCoverage,
QuarantineSource quarantine, OutageSource outage,
CapacitySource capacity, HealthCoverageSource healthCoverage,
QuarantineSource quarantine, OutageSource outage,
LeadSeatSource leadSeats, Map<String, String> leads, String selfTerm,
CoordinationSource coordination, boolean callerIsPrimary) {
return listFleet(workers, sessions, messages, capacity, healthCoverage, LoopHealthSource.none(), quarantine,
outage, leadSeats, leads, selfTerm, coordination, callerIsPrimary);
}
static McpSchema.CallToolResult listFleet(PeerLauncher workers, SessionManager sessions, MessageService messages,
CapacitySource capacity, HealthCoverageSource healthCoverage,
LoopHealthSource loopHealth,
QuarantineSource quarantine, OutageSource outage,
LeadSeatSource leadSeats, Map<String, String> leads, String selfTerm,
CoordinationSource coordination, boolean callerIsPrimary) {
try {
@@ -1656,6 +1706,9 @@ public final class FleetMcp {
Map<String, Object> result = new LinkedHashMap<>();
result.put("leads", leadRows); result.put("members", out);
result.put("healthCoverage", healthCoverage.value().get());
result.put("loopHealth", Map.of(
"statusPoller", loopHealth.statusPoller().get().name(),
"sessionReaper", loopHealth.sessionReaper().get().name()));
// fleetd #439: coordinator/coordinatorView is lead-to-lead coordination state and must
// never reach a worker or an architect -- gate BEFORE assembling it, not after, so the
// key is absent rather than present-and-empty.
@@ -2135,9 +2188,10 @@ public final class FleetMcp {
+ "cannot reliably re-identify: some backends (e.g. opencode) resolve it from the "
+ "member's working directory, which only uniquely identifies a member when it "
+ "was spawned into its own fleetd-provisioned worktree (worktree:true/<slug>); a "
+ "member spawned without one shares its directory with others and never reports "
+ "an id, however long it runs (fleetd #249). An empty 'members' "
+ "means no members are spawned; it says nothing about peers. When capacity "
+ "member spawned without one shares its directory with others and never reports "
+ "an id, however long it runs (fleetd #249). An empty 'members' "
+ "means no members are spawned; it says nothing about peers. 'loopHealth' reports "
+ "the RUNNING, STALLED, or STOPPED state of statusPoller and sessionReaper. When capacity "
+ "facts are configured, a 'capacity' row per profile reports 'free' — the "
+ "slots a fresh fleet_spawn on that profile will actually be granted right "
+ "now (max(0, maxLoad - live)), the same check the spawn gate itself runs. A "
fleetd/src/main/java/dev/ltms/fleet/rest/FleetApp.java
+42 -12
View File
@@ -94,6 +94,7 @@ public final class FleetApp {
// .none() (the honest "feature not wired" view) for every constructor that does not pass one.
private final FleetMcp.QuarantineSource quarantine;
private final FleetMcp.OutageSource outage;
private final FleetMcp.LoopHealthSource loopHealth;
private final ObjectMapper mapper = new ObjectMapper();
/**
@@ -155,7 +156,8 @@ public final class FleetApp {
HttpServlet mcpServlet, CallerResolver auth, Metrics metrics,
Predicate<String> deliverable, Supplier<MemberCredentialPolicyView> memberCredentials) {
this(herdr, memberHerdr, workers, sessions, messages, presence, mcpServlet, auth, metrics,
deliverable, memberCredentials, FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none());
deliverable, memberCredentials, FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(),
FleetMcp.LoopHealthSource.none());
}
/**
@@ -169,8 +171,18 @@ public final class FleetApp {
public FleetApp(HerdrClient herdr, HerdrClient memberHerdr, PeerLauncher workers, SessionManager sessions,
MessageService messages, MemberPresence presence,
HttpServlet mcpServlet, CallerResolver auth, Metrics metrics,
Predicate<String> deliverable, Supplier<MemberCredentialPolicyView> memberCredentials,
FleetMcp.QuarantineSource quarantine, FleetMcp.OutageSource outage) {
Predicate<String> deliverable, Supplier<MemberCredentialPolicyView> memberCredentials,
FleetMcp.QuarantineSource quarantine, FleetMcp.OutageSource outage) {
this(herdr, memberHerdr, workers, sessions, messages, presence, mcpServlet, auth, metrics, deliverable,
memberCredentials, quarantine, outage, FleetMcp.LoopHealthSource.none());
}
public FleetApp(HerdrClient herdr, HerdrClient memberHerdr, PeerLauncher workers, SessionManager sessions,
MessageService messages, MemberPresence presence,
HttpServlet mcpServlet, CallerResolver auth, Metrics metrics,
Predicate<String> deliverable, Supplier<MemberCredentialPolicyView> memberCredentials,
FleetMcp.QuarantineSource quarantine, FleetMcp.OutageSource outage,
FleetMcp.LoopHealthSource loopHealth) {
this.herdr = herdr;
this.memberHerdr = memberHerdr != null ? memberHerdr : herdr;
this.workers = workers;
@@ -183,6 +195,7 @@ public final class FleetApp {
this.memberCredentials = memberCredentials != null ? memberCredentials : MemberCredentialPolicyView::absent;
this.quarantine = quarantine != null ? quarantine : FleetMcp.QuarantineSource.none();
this.outage = outage != null ? outage : FleetMcp.OutageSource.none();
this.loopHealth = loopHealth != null ? loopHealth : FleetMcp.LoopHealthSource.none();
}
/** Wire routes onto a fresh, unstarted Javalin instance. Caller starts it. */
@@ -291,15 +304,19 @@ public final class FleetApp {
* spotted by comparing two numbers by eye.
*/
private void healthz(Context ctx) {
HealthzResponse response = healthzResponse(herdr, memberHerdr, loopHealth);
ctx.status(response.status()).json(response.body());
}
record HealthzResponse(int status, Map<String, Object> body) { }
static HealthzResponse healthzResponse(HerdrClient herdr, HerdrClient memberHerdr,
FleetMcp.LoopHealthSource loopHealth) {
JsonNode pong;
try {
pong = herdr.call("ping");
} catch (HerdrException e) {
ctx.status(503).json(Map.of(
"status", "degraded",
"herdr", "unreachable",
"detail", e.getMessage()));
return;
return degradedResponse("unreachable", e.getMessage(), loopHealth);
}
Map<String, Object> body = new LinkedHashMap<>();
body.put("status", "ok");
@@ -311,11 +328,11 @@ public final class FleetApp {
try {
memberPong = memberHerdr.call("ping");
} catch (HerdrException e) {
ctx.status(503).json(Map.of(
return new HealthzResponse(503, Map.of(
"status", "degraded",
"herdr", "member unreachable",
"detail", e.getMessage()));
return;
"detail", e.getMessage(),
"loopHealth", loopHealthView(loopHealth)));
}
int leadProtocol = pong.path("protocol").asInt();
int memberProtocol = memberPong.path("protocol").asInt();
@@ -326,7 +343,20 @@ public final class FleetApp {
body.put("protocolMismatch", true);
}
}
ctx.status(200).json(body);
body.put("loopHealth", loopHealthView(loopHealth));
return new HealthzResponse(200, body);
}
private static Map<String, String> loopHealthView(FleetMcp.LoopHealthSource loopHealth) {
return Map.of("statusPoller", loopHealth.statusPoller().get().name(),
"sessionReaper", loopHealth.sessionReaper().get().name());
}
private static HealthzResponse degradedResponse(String herdr, String detail,
FleetMcp.LoopHealthSource loopHealth) {
return new HealthzResponse(503, Map.of(
"status", "degraded", "herdr", herdr, "detail", detail,
"loopHealth", loopHealthView(loopHealth)));
}
/**
fleetd/src/test/java/dev/ltms/fleet/mcp/FleetMcpTest.java
+31
View File
@@ -8,6 +8,7 @@ import dev.ltms.fleet.config.FleetConfig;
import dev.ltms.fleet.guard.SubscriptionGuard;
import dev.ltms.fleet.herdr.AgentControl;
import dev.ltms.fleet.herdr.FakeHerdr;
import dev.ltms.fleet.inject.LoopWatchdog;
import dev.ltms.fleet.herdr.PaneLocator;
import dev.ltms.fleet.herdr.WorkspaceControl;
import dev.ltms.fleet.inject.Injector;
@@ -1053,6 +1054,36 @@ class FleetMcpTest {
assertFalse(out.contains("quarantinedForSeconds"), out);
}
@Test
void loopHealthReportsStalledStatusPoller() {
String out = loopHealth(LoopWatchdog.State.STALLED, LoopWatchdog.State.RUNNING);
assertTrue(out.contains("\"statusPoller\":\"STALLED\""),
"fleet_list must report a stalled StatusPoller: " + out);
}
@Test
void loopHealthReportsStoppedSessionReaperAsStopped() {
String out = loopHealth(LoopWatchdog.State.RUNNING, LoopWatchdog.State.STOPPED);
assertTrue(out.contains("\"sessionReaper\":\"STOPPED\""),
"fleet_list must report a deliberately stopped SessionReaper as STOPPED, not an alarm: " + out);
}
@Test
void loopHealthReportsRunningStatusPoller() {
String out = loopHealth(LoopWatchdog.State.RUNNING, LoopWatchdog.State.STOPPED);
assertTrue(out.contains("\"statusPoller\":\"RUNNING\""),
"fleet_list must report a running StatusPoller: " + out);
}
private static String loopHealth(LoopWatchdog.State statusPoller, LoopWatchdog.State sessionReaper) {
FakeHerdr herdr = new FakeHerdr();
return textOf(FleetMcp.listFleet(workerService(herdr, "http://gx00.gw:8000", Set.of("gx00.gw")),
new SessionManager(workerService(herdr, "http://gx00.gw:8000", Set.of("gx00.gw"))), null,
FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
new FleetMcp.LoopHealthSource(() -> statusPoller, () -> sessionReaper),
FleetMcp.QuarantineSource.none(), Map.of(), ""));
}
@Test
void capacityIncludesConfiguredProfileWithoutMembers() {
FakeHerdr h = new FakeHerdr();
fleetd/src/test/java/dev/ltms/fleet/rest/FleetAppTest.java
+24
View File
@@ -8,6 +8,7 @@ import dev.ltms.fleet.herdr.AgentControl;
import dev.ltms.fleet.herdr.FakeHerdr;
import dev.ltms.fleet.herdr.WorkspaceControl;
import dev.ltms.fleet.inject.Injector;
import dev.ltms.fleet.inject.LoopWatchdog;
import dev.ltms.fleet.inject.StatusPoller;
import dev.ltms.fleet.inject.MemberPresence;
import dev.ltms.fleet.msg.InMemoryReplyInbox;
@@ -165,6 +166,29 @@ class FleetAppTest {
assertEquals("degraded", mapper.readTree(res.body()).get("status").asText());
}
@Test
void healthzKeepsOkStatusAndReportsLoopHealthInItsBody() {
FleetMcp.LoopHealthSource loops = new FleetMcp.LoopHealthSource(
() -> LoopWatchdog.State.RUNNING, () -> LoopWatchdog.State.STOPPED);
FleetApp.HealthzResponse ok = FleetApp.healthzResponse(new FakeHerdr(), new FakeHerdr(), loops);
assertEquals(200, ok.status(), "a healthy herdr must keep /healthz at 200 regardless of loop states");
assertEquals(Map.of("statusPoller", "RUNNING", "sessionReaper", "STOPPED"), ok.body().get("loopHealth"),
"the /healthz body must report each loop state without making STOPPED an alarm");
}
@Test
void healthzKeepsDegradedStatusAndReportsLoopHealthInItsBody() {
FleetMcp.LoopHealthSource loops = new FleetMcp.LoopHealthSource(
() -> LoopWatchdog.State.RUNNING, () -> LoopWatchdog.State.STOPPED);
FleetApp.HealthzResponse degraded = FleetApp.healthzResponse(new FakeHerdr().healthy(false),
new FakeHerdr(), loops);
assertEquals(503, degraded.status(),
"an unreachable herdr must keep /healthz at 503 regardless of loop states");
assertEquals(Map.of("statusPoller", "RUNNING", "sessionReaper", "STOPPED"),
degraded.body().get("loopHealth"), "the degraded /healthz body must retain loop states");
}
@Test
void sessionsMapsWorkspaceList() throws Exception {
int port = startHealthy();