fix(stats): keep TopologyDecky.state in sync with docker so ACTIVE DECKIES counts right

Dashboard's ACTIVE DECKIES (active_deckies in get_stats_summary) counts
TopologyDecky rows where state='running'.  No code path was flipping
that state away from the default 'pending', so the count read 0/N
even when every container was running fine — the dashboard was lying.

Two complementary fixes:

1. deploy_topology — after the post-deploy compose ps verification,
   reconcile each TopologyDecky.state from the corresponding base
   container's docker state.  running → 'running'; anything else →
   'failed'.  Reuses the ps_rows already gathered for the
   ACTIVE-vs-DEGRADED status decision; no extra docker hit.

2. apply_add_decky — _materialise_decky_spawn now returns True/False;
   on True the row is updated to state='running' before
   _assert_valid_after.  Catches the case where a decky added via the
   live mutator queue stays at 'pending' indefinitely (the deployer's
   reconcile only runs on a fresh deploy_topology pass).

Existing topology deckies in active topologies will still read as
'pending' until the next deploy_topology runs, since this is
forward-only.  An operator-side fix is to teardown + redeploy or run
the (forthcoming) reconcile-on-startup pass.

decnet/engine/deployer.py

@@ -1005,8 +1005,18 @@ async def deploy_topology(repo, topology_id: str, *, dry_run: bool = False) -> N
         lambda: _compose_ps(compose_path),
     )
     bad: list[str] = []
+    # Build the per-decky state map.  The base container's compose
+    # service name == decky name, which is what we cache on the
+    # TopologyDecky row.  Service containers (named ``<decky>-<svc>``)
+    # don't gate the decky's state — service-level failures are visible
+    # in compose ps separately and don't downgrade the decky as a whole.
+    decky_state_by_name: dict[str, str] = {}
     for row in ps_rows:
         state = str(row.get("State", "")).lower()
+        service_name = str(row.get("Service") or "")
+        if service_name and "-" not in service_name:
+            # Plain decky base; cache its docker state.
+            decky_state_by_name[service_name] = state or "unknown"
         if state and state != "running":
             name = str(row.get("Name") or row.get("Service") or "?")
             exit_code = row.get("ExitCode")
@@ -1015,6 +1025,27 @@ async def deploy_topology(repo, topology_id: str, *, dry_run: bool = False) -> N
                 + (f" (exit={exit_code})" if exit_code not in (None, 0, "") else "")
             )
 
+    # Reconcile each TopologyDecky.state from compose's view.  Without
+    # this, the row stays at the default 'pending' forever and the
+    # dashboard's ACTIVE DECKIES count reads 0/N even when everything's
+    # actually up.
+    for decky in hydrated["deckies"]:
+        cfg = decky.get("decky_config") or {}
+        decky_name = cfg.get("name") or decky.get("name")
+        if not decky_name:
+            continue
+        ds = decky_state_by_name.get(decky_name, "unknown")
+        new_state = "running" if ds == "running" else "failed"
+        try:
+            await repo.update_topology_decky(
+                decky["uuid"], {"state": new_state},
+            )
+        except Exception as exc:  # noqa: BLE001
+            log.warning(
+                "post-deploy state reconcile failed topology=%s decky=%s: %s",
+                topology_id, decky_name, exc,
+            )
+
     if bad:
         reason = "post-deploy check: " + ", ".join(bad[:8]) + (
             f" and {len(bad) - 8} more" if len(bad) > 8 else ""

decnet/mutator/ops.py

@@ -326,16 +326,18 @@ async def _rerender_compose(repo: Any, topology_id: str) -> None:
 
 async def _materialise_decky_spawn(
     repo: Any, topology_id: str, decky_name: str, services: list[str],
-) -> None:
+) -> bool:
     """compose up -d --no-deps --build for one decky (base + services).
 
-    Re-renders compose first so the file lists the new decky.  No-op
+    Re-renders compose first so the file lists the new decky.  Returns
-    when the topology isn't eligible for live materialisation (see
+    True when compose-up reported success, False otherwise (or when
-    :func:`_live_topology_or_none`).  Best-effort: docker failure is
+    the topology isn't eligible for live materialisation — pending
-    logged, not re-raised — DB row is the source of truth.
+    topologies skip and return False so the caller doesn't flip the
+    state to ``running`` based on a no-op).  Best-effort: docker
+    failure is logged, not re-raised — DB row is the source of truth.
     """
     if await _live_topology_or_none(repo, topology_id) is None:
-        return
+        return False
     from decnet.engine.deployer import _topology_compose_path
     await _rerender_compose(repo, topology_id)
     targets = _decky_targets(decky_name, services)
@@ -346,11 +348,13 @@ async def _materialise_decky_spawn(
             compose_file=compose_path,
             label=f"live add_decky topology={topology_id} decky={decky_name}",
         )
+        return True
     except Exception as exc:  # noqa: BLE001
         _log.error(
             "live add_decky: compose up failed topology=%s decky=%s: %s",
             topology_id, decky_name, exc,
         )
+        return False
 
 
 async def _materialise_decky_remove(
@@ -700,7 +704,22 @@ async def apply_add_decky(
     # Live materialisation: spawn the new decky's containers without
     # touching siblings.  Skips on pending / agent-pinned topologies —
     # see _live_topology_or_none.
-    await _materialise_decky_spawn(repo, topology_id, name, services_list)
+    spawned = await _materialise_decky_spawn(
+        repo, topology_id, name, services_list,
+    )
+    # Flip the row's state to 'running' on success so the dashboard's
+    # ACTIVE DECKIES count reflects reality.  Without this the row
+    # stays at the default 'pending' forever; the deployer's full
+    # post-deploy reconcile only runs on a fresh deploy_topology.
+    if spawned:
+        try:
+            await repo.update_topology_decky(decky_uuid, {"state": "running"})
+        except Exception as exc:  # noqa: BLE001
+            _log.warning(
+                "live add_decky: state flip to running failed "
+                "topology=%s decky=%s: %s",
+                topology_id, name, exc,
+            )
     await _assert_valid_after(repo, topology_id)
 
 

tests/mutator/test_ops_materialisation.py

@@ -130,6 +130,23 @@ async def test_add_decky_spawns_base_and_service_containers(repo, stubs):
     assert "newbox-ssh" in args
 
 
+@pytest.mark.anyio
+async def test_add_decky_flips_state_to_running_after_spawn(repo, stubs):
+    """Without this the dashboard's ACTIVE DECKIES count reads 0/N."""
+    tid = await _make_active(repo)
+    lans = await repo.list_lans_for_topology(tid)
+    home_lan = lans[0]["name"]
+
+    await apply_add_decky(repo, tid, {
+        "name": "newrunner",
+        "lan": home_lan,
+        "services": [],
+    })
+    rows = await repo.list_topology_deckies(tid)
+    new = next(r for r in rows if r["name"] == "newrunner")
+    assert new["state"] == "running"
+
+
 @pytest.mark.anyio
 async def test_add_decky_skips_materialisation_when_pending(repo, stubs):
     """Pending topology gets DB write only — deploy_topology will spawn."""