Compare commits
8
Commits
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
334cd068d2 | ||
|
|
f30ea04ab6 | ||
|
|
bb24c77676 | ||
|
|
75d822d0f2 | ||
|
|
dd2b90872a | ||
|
|
5c9bc7eb9c | ||
|
|
5c1d962bf2 | ||
|
|
a3fe1d147c |
+17
-17
@@ -7,7 +7,7 @@ use crate::manifest::Manifest;
|
|||||||
use crate::snapshot;
|
use crate::snapshot;
|
||||||
use crate::sync::{SyncQueue, drain_sync_queue};
|
use crate::sync::{SyncQueue, drain_sync_queue};
|
||||||
use crate::zfs::SystemZfs;
|
use crate::zfs::SystemZfs;
|
||||||
use anyhow::Result;
|
use anyhow::{Context, Result};
|
||||||
use chrono::Utc;
|
use chrono::Utc;
|
||||||
use sysinfo::{ProcessRefreshKind, RefreshKind, System};
|
use sysinfo::{ProcessRefreshKind, RefreshKind, System};
|
||||||
use tokio::time::{interval, Duration};
|
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_dir = cfg.hot.path.clone();
|
||||||
let hot_max_bytes = cfg.hot.max_gb.saturating_mul(1024 * 1024 * 1024);
|
let hot_max_bytes = cfg.hot.max_gb.saturating_mul(1024 * 1024 * 1024);
|
||||||
let blob_root = cluster_cfg.blob_store_root.clone();
|
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,
|
cluster_cfg,
|
||||||
cfg.node.name.clone(),
|
cfg.node.name.clone(),
|
||||||
hot_dir,
|
hot_dir,
|
||||||
@@ -44,21 +51,14 @@ pub async fn run(cfg: Config, mut manifest: Manifest) -> Result<()> {
|
|||||||
blob_root,
|
blob_root,
|
||||||
)
|
)
|
||||||
.await
|
.await
|
||||||
{
|
.context("starting cluster services")?;
|
||||||
Ok(svc) => {
|
tracing::info!(
|
||||||
tracing::info!(
|
rpc_enabled = svc.rpc_enabled(),
|
||||||
rpc_enabled = svc.rpc_enabled(),
|
blob_store_enabled = svc.blob_store_enabled(),
|
||||||
blob_store_enabled = svc.blob_store_enabled(),
|
zone = %cluster_cfg.zone,
|
||||||
zone = %cluster_cfg.zone,
|
"cluster services online"
|
||||||
"cluster services online"
|
);
|
||||||
);
|
Some(svc)
|
||||||
Some(svc)
|
|
||||||
}
|
|
||||||
Err(e) => {
|
|
||||||
tracing::error!(error = %e, "cluster services failed to start; continuing without cluster");
|
|
||||||
None
|
|
||||||
}
|
|
||||||
}
|
|
||||||
}
|
}
|
||||||
None => {
|
None => {
|
||||||
tracing::info!("no [cluster] section in config; running standalone");
|
tracing::info!("no [cluster] section in config; running standalone");
|
||||||
|
|||||||
@@ -91,9 +91,30 @@ impl V2State {
|
|||||||
.join("aggregator-sessions.json");
|
.join("aggregator-sessions.json");
|
||||||
let sessions = SessionStore::load(sessions_path)
|
let sessions = SessionStore::load(sessions_path)
|
||||||
.map_err(|e| anyhow::anyhow!("loading session store: {e}"))?;
|
.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 {
|
Ok(Self {
|
||||||
aggregator_name: cfg.node.name.clone(),
|
aggregator_name: cfg.node.name.clone(),
|
||||||
peers: cluster.peers.clone(),
|
peers,
|
||||||
client: std::sync::Arc::new(client),
|
client: std::sync::Arc::new(client),
|
||||||
default_rpc_port_offset: 1,
|
default_rpc_port_offset: 1,
|
||||||
api_token: cfg.api_token.clone(),
|
api_token: cfg.api_token.clone(),
|
||||||
|
|||||||
@@ -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"}
|
||||||
@@ -6,13 +6,15 @@ import react from '@vitejs/plugin-react';
|
|||||||
// During local dev the daemon proxies /api/v2/* on :7700 so
|
// During local dev the daemon proxies /api/v2/* on :7700 so
|
||||||
// `vite dev` on :5173 can hit it via server.proxy.
|
// `vite dev` on :5173 can hit it via server.proxy.
|
||||||
export default defineConfig({
|
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()],
|
plugins: [react()],
|
||||||
server: {
|
server: {
|
||||||
port: 5173,
|
port: 5173,
|
||||||
|
|||||||
Reference in New Issue
Block a user