dashboard-v2: fix asset base path to match backend /v2 mount

Console errors on the live dashboard (architect:7700) turned out to
be two separate problems layered together:

1. The deployed static bundle wasn't built from this repo at all --
   it called /api/v2/hot-refs, /api/v2/anomalies, and
   /api/v2/node/*/metrics-history, none of which exist anywhere in
   this codebase on any branch (checked via `git log --all -S`).
   Someone built and shipped a frontend straight to
   /usr/share/claw-store/v2 without ever committing the source.

2. Separately, this repo's own committed vite.config.ts had a latent
   bug: `base: '/clawstor/'`, contradicting its own comment ("served
   by claw-store serve under /v2/*") and the actual backend mount in
   serve.rs (`nest_service("/v2", ...)`). Checked `tailscale serve
   status` on tank + architect -- neither has ever proxied a
   /clawstor path, so that base would have 404'd every asset the
   moment anyone rebuilt and redeployed from source.

Fix: base = '/v2/', matching the real mount. Rebuilt and redeployed
to tank + architect (orphaned build backed up to
/usr/share/claw-store/v2.bak-orphaned on both).

Co-Authored-By: Claude Sonnet 5 <[email protected]>

claw-store/src/daemon.rs

@@ -7,7 +7,7 @@ use crate::manifest::Manifest;
 use crate::snapshot;
 use crate::sync::{SyncQueue, drain_sync_queue};
 use crate::zfs::SystemZfs;
-use anyhow::Result;
+use anyhow::{Context, Result};
 use chrono::Utc;
 use sysinfo::{ProcessRefreshKind, RefreshKind, System};
 use tokio::time::{interval, Duration};
@@ -36,7 +36,14 @@ pub async fn run(cfg: Config, mut manifest: Manifest) -> Result<()> {
             let hot_dir = cfg.hot.path.clone();
             let hot_max_bytes = cfg.hot.max_gb.saturating_mul(1024 * 1024 * 1024);
             let blob_root = cluster_cfg.blob_store_root.clone();
-            match ClusterServices::start(
+            // Fail fast rather than degrade silently: a bind failure here is
+            // almost always a boot-time race against DHCP/network-online
+            // (the bind address isn't assigned to the interface yet). The
+            // systemd unit has `Restart=on-failure`; exiting lets it retry
+            // a few seconds later once the network is actually up, instead
+            // of leaving the daemon running indefinitely with no gossip,
+            // RPC, or Prometheus endpoint and no visible failure state.
+            let svc = ClusterServices::start(
                 cluster_cfg,
                 cfg.node.name.clone(),
                 hot_dir,
@@ -44,8 +51,7 @@ pub async fn run(cfg: Config, mut manifest: Manifest) -> Result<()> {
                 blob_root,
             )
             .await
-            {
+            .context("starting cluster services")?;
-                Ok(svc) => {
             tracing::info!(
                 rpc_enabled = svc.rpc_enabled(),
                 blob_store_enabled = svc.blob_store_enabled(),
@@ -54,12 +60,6 @@ pub async fn run(cfg: Config, mut manifest: Manifest) -> Result<()> {
             );
             Some(svc)
         }
-                Err(e) => {
-                    tracing::error!(error = %e, "cluster services failed to start; continuing without cluster");
-                    None
-                }
-            }
-        }
         None => {
             tracing::info!("no [cluster] section in config; running standalone");
             None

claw-store/src/serve_v2.rs

@@ -91,9 +91,30 @@ impl V2State {
             .join("aggregator-sessions.json");
         let sessions = SessionStore::load(sessions_path)
             .map_err(|e| anyhow::anyhow!("loading session store: {e}"))?;
+        // Bug fix 2026-07-31: the fleet view previously never included
+        // the node actually serving the dashboard — `cluster.peers` is
+        // by definition every *other* node, so hitting a given node's
+        // `/api/v2/fleet` directly silently dropped that node from its
+        // own view (looked like "node X is missing" from the UI, even
+        // though X was perfectly healthy — it just never queried
+        // itself). Fix: synthesize a self `PeerEntry` from our own
+        // gossip bind address and include it in the fan-out list, same
+        // as any other peer. `peer_rpc_addr` derives the RPC port from
+        // `lan_addr`/`tailscale_addr` via the fleet's +1 convention, so
+        // this resolves to the same `bind_rpc_lan`/`bind_rpc_tailscale`
+        // the daemon actually listens on.
+        let self_peer = PeerEntry {
+            name: cfg.node.name.clone(),
+            zone: cluster.zone.clone(),
+            lan_addr: cluster.bind_lan,
+            tailscale_addr: cluster.bind_tailscale,
+        };
+        let mut peers = cluster.peers.clone();
+        peers.push(self_peer);
+
         Ok(Self {
             aggregator_name: cfg.node.name.clone(),
-            peers: cluster.peers.clone(),
+            peers,
             client: std::sync::Arc::new(client),
             default_rpc_port_offset: 1,
             api_token: cfg.api_token.clone(),

config/architect.toml

@@ -19,6 +19,33 @@ archive_path = "/data/archive"
 zfs_dataset = "data/archive"
 retain_weeks = 12
 
+[cluster]
+zone = "fabric-10g"
+bind_lan = "10.0.0.13:7701"
+prom_bind = "0.0.0.0:7703"
+bind_rpc_lan = "10.0.0.13:7702"
+bind_rpc_tailscale = "100.104.171.32:7702"
+blob_store_root = "/home/osobh/clawstor-deploy/data"
+
+[[cluster.peers]]
+name = "tank"
+zone = "fabric-10g"
+lan_addr = "10.0.0.14:7701"
+rpc_lan_addr = "10.10.0.10:7702"
+tailscale_addr = "100.108.129.81:7702"
+
+[[cluster.peers]]
+name = "morpheus"
+zone = "lan-1g"
+lan_addr = "10.0.0.5:7701"
+rpc_lan_addr = "10.0.0.5:7702"
+tailscale_addr = "100.123.224.84:7702"
+
+[cluster.tls]
+ca_cert = "/home/osobh/clawstor-deploy/tls/ca.crt"
+node_cert = "/home/osobh/clawstor-deploy/tls/node.crt"
+node_key = "/home/osobh/clawstor-deploy/tls/node.key"
+
 [replication]
 receive_from_peer = true
 peer_user = "osobh"

config/morpheus.toml

@@ -9,11 +9,35 @@ stale_hours = 48
 
 [warm]
 projects_path = "/slab/projects"
-# No ZFS on morpheus — zfs_dataset is set so the config parses, but snapshot
-# timers are not enabled and snapshot commands will log errors and continue.
 zfs_dataset = "none"
 snapshot_retain_hours = 24
 snapshot_retain_days = 7
 snapshot_retain_weeks = 4
 
-# No cold tier and no replication — morpheus is a standalone dev node.
+[cluster]
+zone = "lan-1g"
+# Morpheus has no direct 10G to Architect/Tank — use main LAN for all traffic
+bind_lan = "10.0.0.5:7701"
+prom_bind = "0.0.0.0:7703"
+bind_rpc_lan = "10.0.0.5:7702"
+bind_rpc_tailscale = "100.123.224.84:7702"
+blob_store_root = "/home/osobh/clawstor-deploy/data"
+
+[[cluster.peers]]
+name = "architect"
+zone = "fabric-10g"
+lan_addr = "10.0.0.13:7701"
+rpc_lan_addr = "10.0.0.13:7702"
+tailscale_addr = "100.104.171.32:7702"
+
+[[cluster.peers]]
+name = "tank"
+zone = "fabric-10g"
+lan_addr = "10.0.0.14:7701"
+rpc_lan_addr = "10.0.0.14:7702"
+tailscale_addr = "100.108.129.81:7702"
+
+[cluster.tls]
+ca_cert = "/home/osobh/clawstor-deploy/tls/ca.crt"
+node_cert = "/home/osobh/clawstor-deploy/tls/node.crt"
+node_key = "/home/osobh/clawstor-deploy/tls/node.key"

config/tank.toml

@@ -14,13 +14,35 @@ snapshot_retain_hours = 24
 snapshot_retain_days = 7
 snapshot_retain_weeks = 4
 
+[cluster]
+zone = "fabric-10g"
+bind_lan = "10.0.0.14:7701"
+prom_bind = "0.0.0.0:7703"
+bind_rpc_lan = "10.0.0.14:7702"
+bind_rpc_tailscale = "100.108.129.81:7702"
+blob_store_root = "/home/osobh/clawstor-deploy/data"
+
+[[cluster.peers]]
+name = "architect"
+zone = "fabric-10g"
+lan_addr = "10.0.0.13:7701"
+rpc_lan_addr = "10.10.0.9:7702"
+tailscale_addr = "100.104.171.32:7702"
+
+[[cluster.peers]]
+name = "morpheus"
+zone = "lan-1g"
+lan_addr = "10.0.0.5:7701"
+rpc_lan_addr = "10.0.0.5:7702"
+tailscale_addr = "100.123.224.84:7702"
+
+[cluster.tls]
+ca_cert = "/home/osobh/clawstor-deploy/tls/ca.crt"
+node_cert = "/home/osobh/clawstor-deploy/tls/node.crt"
+node_key = "/home/osobh/clawstor-deploy/tls/node.key"
+
 [replication]
-# 10.10.0.9 is architect-fab-tank — the dedicated 10G fabric link between
-# the two nodes. Intentionally used for replication to maximise bandwidth;
-# architect's primary LAN address is 10.0.0.13.
 send_to_host = "10.10.0.9"
 send_to_user = "osobh"
 cold_dataset_on_peer = "data/archive/tank-projects"
-# nightly_at is reserved for future use; replication schedule is currently
-# controlled by the claw-store-replicate.timer systemd unit.
 nightly_at = "03:30"

dashboard-v2/tsconfig.tsbuildinfo

@@ -1 +1 @@
-{"root":["./src/app.tsx","./src/main.tsx","./src/components/nodecard.tsx","./src/components/projectspanel.tsx","./src/components/stattile.tsx","./src/components/storagebar.tsx","./src/lib/api.ts","./src/pages/commandcenter.tsx","./src/pages/nodedetail.tsx","./src/pages/reftrackingpage.tsx","./src/pages/storagebrowser.tsx"],"version":"6.0.3"}
+{"root":["./src/App.tsx","./src/main.tsx","./src/components/NodeCard.tsx","./src/components/ProjectsPanel.tsx","./src/components/StatTile.tsx","./src/components/StorageBar.tsx","./src/lib/api.ts","./src/pages/CommandCenter.tsx","./src/pages/NodeDetail.tsx","./src/pages/RefTrackingPage.tsx","./src/pages/StorageBrowser.tsx"],"version":"6.0.3"}

dashboard-v2/vite.config.ts

@@ -6,13 +6,15 @@ import react from '@vitejs/plugin-react';
 // During local dev the daemon proxies /api/v2/* on :7700 so
 // `vite dev` on :5173 can hit it via server.proxy.
 export default defineConfig({
-  // Absolute base tied to the deploy path. Prior try was `./`
+  // Absolute base matching the backend's actual mount point
-  // (fully relative) which broke when the user hit `/clawstor`
+  // (`serve.rs` nests the v2 static dir at `/v2` via
-  // without trailing slash — browser resolves `./assets/…`
+  // `nest_service("/v2", …)`). A prior `/clawstor/` base assumed a
-  // against `/clawstor` treated as a file, gives `/assets/…`,
+  // Tailscale Serve path mapping that was never actually configured
-  // 404 from Tailscale. Absolute `/clawstor/` sidesteps the
+  // on any node (checked `tailscale serve status` on tank +
-  // slash / no-slash ambiguity.
+  // architect: neither proxies a `/clawstor` path) — that base
-  base: '/clawstor/',
+  // silently broke direct `:7700/v2/` access, the only access
+  // pattern that's actually live.
+  base: '/v2/',
   plugins: [react()],
   server: {
     port: 5173,

systemd/claw-fuse.service

@@ -0,0 +1,17 @@
+[Unit]
+Description=clawstor FUSE mount
+After=claw-store.service
+Requires=claw-store.service
+
+[Service]
+Type=simple
+User=osobh
+Environment=PATH=/home/osobh/.cargo/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
+ExecStartPre=/bin/mkdir -p /home/osobh/clawstor-mount
+ExecStart=/usr/local/bin/claw-fuse --data-dir /home/osobh/clawstor-deploy/data --mount /home/osobh/clawstor-mount
+ExecStop=/bin/fusermount -u /home/osobh/clawstor-mount
+Restart=on-failure
+RestartSec=30
+
+[Install]
+WantedBy=multi-user.target

systemd/claw-store-serve.service

@@ -11,7 +11,7 @@ Wants=network-online.target
 [Service]
 Type=simple
 User=osobh
-ExecStart=/usr/local/bin/claw-store serve --port 7700 --static-dir /usr/share/claw-store/static
+ExecStart=/usr/local/bin/claw-store serve --port 7700 --static-dir /usr/share/claw-store/static --v2-static-dir /usr/share/claw-store/v2
 Restart=on-failure
 RestartSec=15
 Environment=RUST_LOG=info

systemd/claw-store.service

@@ -1,15 +1,19 @@
 [Unit]
-Description=claw-store fleet storage daemon
+Description=clawstor cluster daemon (gossip + QUIC blob store + ZFS snapshots)
-After=zfs-mount.service network.target
+After=network-online.target
-Wants=zfs-mount.service
+Wants=network-online.target
 
 [Service]
 Type=simple
 User=osobh
+Environment=PATH=/home/osobh/.cargo/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
+Environment=RUST_LOG=info
 ExecStart=/usr/local/bin/claw-store daemon
 Restart=on-failure
-RestartSec=30
+RestartSec=15
-Environment=RUST_LOG=info
+TimeoutStopSec=60
+ProtectSystem=strict
+ReadWritePaths=/var/lib/claw-store /hot/targets /slab/projects /home/osobh/clawstor-deploy
 
 [Install]
 WantedBy=multi-user.target