aboutsummaryrefslogtreecommitdiff
path: root/src/rpc/system.rs
blob: f4fdcace44cabfd1b05029bf321e7427d1e63790 (plain) (blame)
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
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
//! Module containing structs related to membership management
use std::collections::HashMap;
use std::io::{Read, Write};
use std::net::{IpAddr, SocketAddr};
use std::path::{Path, PathBuf};
use std::sync::{Arc, RwLock};
use std::time::{Duration, Instant};

use arc_swap::ArcSwapOption;
use async_trait::async_trait;
use futures::join;
use serde::{Deserialize, Serialize};
use sodiumoxide::crypto::sign::ed25519;
use tokio::select;
use tokio::sync::watch;
use tokio::sync::Mutex;

use garage_net::endpoint::{Endpoint, EndpointHandler};
use garage_net::message::*;
use garage_net::peering::PeeringManager;
use garage_net::util::parse_and_resolve_peer_addr_async;
use garage_net::{NetApp, NetworkKey, NodeID, NodeKey};

#[cfg(feature = "kubernetes-discovery")]
use garage_util::config::KubernetesDiscoveryConfig;
use garage_util::config::{Config, DataDirEnum};
use garage_util::data::*;
use garage_util::error::*;
use garage_util::persister::Persister;
use garage_util::time::*;

#[cfg(feature = "consul-discovery")]
use crate::consul::ConsulDiscovery;
#[cfg(feature = "kubernetes-discovery")]
use crate::kubernetes::*;
use crate::layout::*;
use crate::replication_mode::*;
use crate::ring::*;
use crate::rpc_helper::*;

use crate::system_metrics::*;

const DISCOVERY_INTERVAL: Duration = Duration::from_secs(60);
const STATUS_EXCHANGE_INTERVAL: Duration = Duration::from_secs(10);

/// Version tag used for version check upon Netapp connection.
/// Cluster nodes with different version tags are deemed
/// incompatible and will refuse to connect.
pub const GARAGE_VERSION_TAG: u64 = 0x6761726167650008; // garage 0x0008

/// RPC endpoint used for calls related to membership
pub const SYSTEM_RPC_PATH: &str = "garage_rpc/membership.rs/SystemRpc";

/// RPC messages related to membership
#[derive(Debug, Serialize, Deserialize, Clone)]
pub enum SystemRpc {
	/// Response to successfull advertisements
	Ok,
	/// Request to connect to a specific node (in <pubkey>@<host>:<port> format, pubkey = full-length node ID)
	Connect(String),
	/// Ask other node its cluster layout. Answered with AdvertiseClusterLayout
	PullClusterLayout,
	/// Advertise Garage status. Answered with another AdvertiseStatus.
	/// Exchanged with every node on a regular basis.
	AdvertiseStatus(NodeStatus),
	/// Advertisement of cluster layout. Sent spontanously or in response to PullClusterLayout
	AdvertiseClusterLayout(ClusterLayout),
	/// Get known nodes states
	GetKnownNodes,
	/// Return known nodes
	ReturnKnownNodes(Vec<KnownNodeInfo>),
}

impl Rpc for SystemRpc {
	type Response = Result<SystemRpc, Error>;
}

#[derive(Serialize, Deserialize)]
pub struct PeerList(Vec<(Uuid, SocketAddr)>);
impl garage_util::migrate::InitialFormat for PeerList {}

/// This node's membership manager
pub struct System {
	/// The id of this node
	pub id: Uuid,

	persist_cluster_layout: Persister<ClusterLayout>,
	persist_peer_list: Persister<PeerList>,

	pub(crate) local_status: RwLock<NodeStatus>,
	node_status: RwLock<HashMap<Uuid, (u64, NodeStatus)>>,

	pub netapp: Arc<NetApp>,
	peering: Arc<PeeringManager>,
	pub rpc: RpcHelper,

	system_endpoint: Arc<Endpoint<SystemRpc, System>>,

	rpc_listen_addr: SocketAddr,
	rpc_public_addr: Option<SocketAddr>,
	bootstrap_peers: Vec<String>,

	#[cfg(feature = "consul-discovery")]
	consul_discovery: Option<ConsulDiscovery>,
	#[cfg(feature = "kubernetes-discovery")]
	kubernetes_discovery: Option<KubernetesDiscoveryConfig>,

	metrics: ArcSwapOption<SystemMetrics>,

	replication_mode: ReplicationMode,
	pub(crate) replication_factor: usize,

	/// The ring
	pub ring: watch::Receiver<Arc<Ring>>,
	update_ring: Mutex<watch::Sender<Arc<Ring>>>,

	/// Path to metadata directory
	pub metadata_dir: PathBuf,
	/// Path to data directory
	pub data_dir: DataDirEnum,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct NodeStatus {
	/// Hostname of the node
	pub hostname: String,

	/// Replication factor configured on the node
	pub replication_factor: usize,
	/// Cluster layout version
	pub cluster_layout_version: u64,
	/// Hash of cluster layout staging data
	pub cluster_layout_staging_hash: Hash,

	/// Disk usage on partition containing metadata directory (tuple: `(avail, total)`)
	#[serde(default)]
	pub meta_disk_avail: Option<(u64, u64)>,
	/// Disk usage on partition containing data directory (tuple: `(avail, total)`)
	#[serde(default)]
	pub data_disk_avail: Option<(u64, u64)>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct KnownNodeInfo {
	pub id: Uuid,
	pub addr: SocketAddr,
	pub is_up: bool,
	pub last_seen_secs_ago: Option<u64>,
	pub status: NodeStatus,
}

#[derive(Debug, Clone, Copy)]
pub struct ClusterHealth {
	/// The current health status of the cluster (see below)
	pub status: ClusterHealthStatus,
	/// Number of nodes already seen once in the cluster
	pub known_nodes: usize,
	/// Number of nodes currently connected
	pub connected_nodes: usize,
	/// Number of storage nodes declared in the current layout
	pub storage_nodes: usize,
	/// Number of storage nodes currently connected
	pub storage_nodes_ok: usize,
	/// Number of partitions in the layout
	pub partitions: usize,
	/// Number of partitions for which we have a quorum of connected nodes
	pub partitions_quorum: usize,
	/// Number of partitions for which all storage nodes are connected
	pub partitions_all_ok: usize,
}

#[derive(Debug, Clone, Copy, Eq, PartialEq)]
pub enum ClusterHealthStatus {
	/// All nodes are available
	Healthy,
	/// Some storage nodes are unavailable, but quorum is stil
	/// achieved for all partitions
	Degraded,
	/// Quorum is not available for some partitions
	Unavailable,
}

pub fn read_node_id(metadata_dir: &Path) -> Result<NodeID, Error> {
	let mut pubkey_file = metadata_dir.to_path_buf();
	pubkey_file.push("node_key.pub");

	let mut f = std::fs::File::open(pubkey_file.as_path())?;
	let mut d = vec![];
	f.read_to_end(&mut d)?;
	if d.len() != 32 {
		return Err(Error::Message("Corrupt node_key.pub file".to_string()));
	}

	let mut key = [0u8; 32];
	key.copy_from_slice(&d[..]);
	Ok(NodeID::from_slice(&key[..]).unwrap())
}

pub fn gen_node_key(metadata_dir: &Path) -> Result<NodeKey, Error> {
	let mut key_file = metadata_dir.to_path_buf();
	key_file.push("node_key");
	if key_file.as_path().exists() {
		let mut f = std::fs::File::open(key_file.as_path())?;
		let mut d = vec![];
		f.read_to_end(&mut d)?;
		if d.len() != 64 {
			return Err(Error::Message("Corrupt node_key file".to_string()));
		}

		let mut key = [0u8; 64];
		key.copy_from_slice(&d[..]);
		Ok(NodeKey::from_slice(&key[..]).unwrap())
	} else {
		if !metadata_dir.exists() {
			info!("Metadata directory does not exist, creating it.");
			std::fs::create_dir(metadata_dir)?;
		}

		info!("Generating new node key pair.");
		let (pubkey, key) = ed25519::gen_keypair();

		{
			use std::os::unix::fs::PermissionsExt;
			let mut f = std::fs::File::create(key_file.as_path())?;
			let mut perm = f.metadata()?.permissions();
			perm.set_mode(0o600);
			std::fs::set_permissions(key_file.as_path(), perm)?;
			f.write_all(&key[..])?;
		}

		{
			let mut pubkey_file = metadata_dir.to_path_buf();
			pubkey_file.push("node_key.pub");
			let mut f2 = std::fs::File::create(pubkey_file.as_path())?;
			f2.write_all(&pubkey[..])?;
		}

		Ok(key)
	}
}

impl System {
	/// Create this node's membership manager
	pub fn new(
		network_key: NetworkKey,
		replication_mode: ReplicationMode,
		config: &Config,
	) -> Result<Arc<Self>, Error> {
		let replication_factor = replication_mode.replication_factor();

		let node_key =
			gen_node_key(&config.metadata_dir).expect("Unable to read or generate node ID");
		info!(
			"Node ID of this node: {}",
			hex::encode(&node_key.public_key()[..8])
		);

		let persist_cluster_layout: Persister<ClusterLayout> =
			Persister::new(&config.metadata_dir, "cluster_layout");
		let persist_peer_list = Persister::new(&config.metadata_dir, "peer_list");

		let cluster_layout = match persist_cluster_layout.load() {
			Ok(x) => {
				if x.replication_factor != replication_factor {
					return Err(Error::Message(format!(
						"Prevous cluster layout has replication factor {}, which is different than the one specified in the config file ({}). The previous cluster layout can be purged, if you know what you are doing, simply by deleting the `cluster_layout` file in your metadata directory.",
						x.replication_factor,
						replication_factor
					)));
				}
				x
			}
			Err(e) => {
				info!(
					"No valid previous cluster layout stored ({}), starting fresh.",
					e
				);
				ClusterLayout::new(replication_factor)
			}
		};

		let mut local_status = NodeStatus::initial(replication_factor, &cluster_layout);
		local_status.update_disk_usage(&config.metadata_dir, &config.data_dir);

		let ring = Ring::new(cluster_layout, replication_factor);
		let (update_ring, ring) = watch::channel(Arc::new(ring));

		let rpc_public_addr = match &config.rpc_public_addr {
			Some(a_str) => {
				use std::net::ToSocketAddrs;
				match a_str.to_socket_addrs() {
					Err(e) => {
						error!(
							"Cannot resolve rpc_public_addr {} from config file: {}.",
							a_str, e
						);
						None
					}
					Ok(a) => {
						let a = a.collect::<Vec<_>>();
						if a.is_empty() {
							error!("rpc_public_addr {} resolve to no known IP address", a_str);
						}
						if a.len() > 1 {
							warn!("Multiple possible resolutions for rpc_public_addr: {:?}. Taking the first one.", a);
						}
						a.into_iter().next()
					}
				}
			}
			None => {
				let addr =
					get_default_ip().map(|ip| SocketAddr::new(ip, config.rpc_bind_addr.port()));
				if let Some(a) = addr {
					warn!("Using autodetected rpc_public_addr: {}. Consider specifying it explicitly in configuration file if possible.", a);
				}
				addr
			}
		};
		if rpc_public_addr.is_none() {
			warn!("This Garage node does not know its publicly reachable RPC address, this might hamper intra-cluster communication.");
		}

		let bind_outgoing_to = Some(config)
			.filter(|x| x.rpc_bind_outgoing)
			.map(|x| x.rpc_bind_addr.ip());
		let netapp = NetApp::new(GARAGE_VERSION_TAG, network_key, node_key, bind_outgoing_to);
		let peering = PeeringManager::new(netapp.clone(), vec![], rpc_public_addr);
		if let Some(ping_timeout) = config.rpc_ping_timeout_msec {
			peering.set_ping_timeout_millis(ping_timeout);
		}

		let system_endpoint = netapp.endpoint(SYSTEM_RPC_PATH.into());

		#[cfg(feature = "consul-discovery")]
		let consul_discovery = match &config.consul_discovery {
			Some(cfg) => Some(
				ConsulDiscovery::new(cfg.clone())
					.ok_or_message("Invalid Consul discovery configuration")?,
			),
			None => None,
		};
		#[cfg(not(feature = "consul-discovery"))]
		if config.consul_discovery.is_some() {
			warn!("Consul discovery is not enabled in this build.");
		}

		#[cfg(not(feature = "kubernetes-discovery"))]
		if config.kubernetes_discovery.is_some() {
			warn!("Kubernetes discovery is not enabled in this build.");
		}

		let sys = Arc::new(System {
			id: netapp.id.into(),
			persist_cluster_layout,
			persist_peer_list,
			local_status: RwLock::new(local_status),
			node_status: RwLock::new(HashMap::new()),
			netapp: netapp.clone(),
			peering: peering.clone(),
			rpc: RpcHelper::new(
				netapp.id.into(),
				peering,
				ring.clone(),
				config.rpc_timeout_msec.map(Duration::from_millis),
			),
			system_endpoint,
			replication_mode,
			replication_factor,
			rpc_listen_addr: config.rpc_bind_addr,
			rpc_public_addr,
			bootstrap_peers: config.bootstrap_peers.clone(),
			#[cfg(feature = "consul-discovery")]
			consul_discovery,
			#[cfg(feature = "kubernetes-discovery")]
			kubernetes_discovery: config.kubernetes_discovery.clone(),
			metrics: ArcSwapOption::new(None),

			ring,
			update_ring: Mutex::new(update_ring),
			metadata_dir: config.metadata_dir.clone(),
			data_dir: config.data_dir.clone(),
		});

		sys.system_endpoint.set_handler(sys.clone());

		let metrics = SystemMetrics::new(sys.clone());
		sys.metrics.store(Some(Arc::new(metrics)));

		Ok(sys)
	}

	/// Perform bootstraping, starting the ping loop
	pub async fn run(self: Arc<Self>, must_exit: watch::Receiver<bool>) {
		join!(
			self.netapp.clone().listen(
				self.rpc_listen_addr,
				self.rpc_public_addr,
				must_exit.clone()
			),
			self.peering.clone().run(must_exit.clone()),
			self.discovery_loop(must_exit.clone()),
			self.status_exchange_loop(must_exit.clone()),
		);
	}

	pub fn cleanup(&self) {
		// Break reference cycle
		self.metrics.store(None);
	}

	// ---- Administrative operations (directly available and
	//      also available through RPC) ----

	pub fn get_known_nodes(&self) -> Vec<KnownNodeInfo> {
		let node_status = self.node_status.read().unwrap();
		let known_nodes = self
			.peering
			.get_peer_list()
			.iter()
			.map(|n| KnownNodeInfo {
				id: n.id.into(),
				addr: n.addr,
				is_up: n.is_up(),
				last_seen_secs_ago: n
					.last_seen
					.map(|t| (Instant::now().saturating_duration_since(t)).as_secs()),
				status: node_status
					.get(&n.id.into())
					.cloned()
					.map(|(_, st)| st)
					.unwrap_or_else(NodeStatus::unknown),
			})
			.collect::<Vec<_>>();
		known_nodes
	}

	pub fn get_cluster_layout(&self) -> ClusterLayout {
		self.ring.borrow().layout.clone()
	}

	pub async fn update_cluster_layout(
		self: &Arc<Self>,
		layout: &ClusterLayout,
	) -> Result<(), Error> {
		self.handle_advertise_cluster_layout(layout).await?;
		Ok(())
	}

	pub async fn connect(&self, node: &str) -> Result<(), Error> {
		let (pubkey, addrs) = parse_and_resolve_peer_addr_async(node)
			.await
			.ok_or_else(|| {
				Error::Message(format!(
					"Unable to parse or resolve node specification: {}",
					node
				))
			})?;
		let mut errors = vec![];
		for addr in addrs.iter() {
			match self.netapp.clone().try_connect(*addr, pubkey).await {
				Ok(()) => return Ok(()),
				Err(e) => {
					errors.push((
						*addr,
						Error::Message(connect_error_message(*addr, pubkey, e)),
					));
				}
			}
		}
		if errors.len() == 1 {
			Err(Error::Message(errors[0].1.to_string()))
		} else {
			Err(Error::Message(format!("{:?}", errors)))
		}
	}

	pub fn health(&self) -> ClusterHealth {
		let ring: Arc<_> = self.ring.borrow().clone();
		let quorum = self.replication_mode.write_quorum();
		let replication_factor = self.replication_factor;

		let nodes = self
			.get_known_nodes()
			.into_iter()
			.map(|n| (n.id, n))
			.collect::<HashMap<Uuid, _>>();
		let connected_nodes = nodes.iter().filter(|(_, n)| n.is_up).count();

		let storage_nodes = ring
			.layout
			.roles
			.items()
			.iter()
			.filter(|(_, _, v)| matches!(v, NodeRoleV(Some(r)) if r.capacity.is_some()))
			.collect::<Vec<_>>();
		let storage_nodes_ok = storage_nodes
			.iter()
			.filter(|(x, _, _)| nodes.get(x).map(|n| n.is_up).unwrap_or(false))
			.count();

		let partitions = ring.partitions();
		let partitions_n_up = partitions
			.iter()
			.map(|(_, h)| {
				let pn = ring.get_nodes(h, ring.replication_factor);
				pn.iter()
					.filter(|x| nodes.get(x).map(|n| n.is_up).unwrap_or(false))
					.count()
			})
			.collect::<Vec<usize>>();
		let partitions_all_ok = partitions_n_up
			.iter()
			.filter(|c| **c == replication_factor)
			.count();
		let partitions_quorum = partitions_n_up.iter().filter(|c| **c >= quorum).count();

		let status =
			if partitions_quorum == partitions.len() && storage_nodes_ok == storage_nodes.len() {
				ClusterHealthStatus::Healthy
			} else if partitions_quorum == partitions.len() {
				ClusterHealthStatus::Degraded
			} else {
				ClusterHealthStatus::Unavailable
			};

		ClusterHealth {
			status,
			known_nodes: nodes.len(),
			connected_nodes,
			storage_nodes: storage_nodes.len(),
			storage_nodes_ok,
			partitions: partitions.len(),
			partitions_quorum,
			partitions_all_ok,
		}
	}

	// ---- INTERNALS ----

	#[cfg(feature = "consul-discovery")]
	async fn advertise_to_consul(self: Arc<Self>) {
		let c = match &self.consul_discovery {
			Some(c) => c,
			_ => return,
		};

		let rpc_public_addr = match self.rpc_public_addr {
			Some(addr) => addr,
			None => {
				warn!("Not advertising to Consul because rpc_public_addr is not defined in config file and could not be autodetected.");
				return;
			}
		};

		let hostname = self.local_status.read().unwrap().hostname.clone();
		if let Err(e) = c
			.publish_consul_service(self.netapp.id, &hostname, rpc_public_addr)
			.await
		{
			error!("Error while publishing Consul service: {}", e);
		}
	}

	#[cfg(feature = "kubernetes-discovery")]
	async fn advertise_to_kubernetes(self: Arc<Self>) {
		let k = match &self.kubernetes_discovery {
			Some(k) => k,
			_ => return,
		};

		let rpc_public_addr = match self.rpc_public_addr {
			Some(addr) => addr,
			None => {
				warn!("Not advertising to Kubernetes because rpc_public_addr is not defined in config file and could not be autodetected.");
				return;
			}
		};

		let hostname = self.local_status.read().unwrap().hostname.clone();
		if let Err(e) = publish_kubernetes_node(k, self.netapp.id, &hostname, rpc_public_addr).await
		{
			error!("Error while publishing node to Kubernetes: {}", e);
		}
	}

	/// Save network configuration to disc
	async fn save_cluster_layout(&self) -> Result<(), Error> {
		let ring: Arc<Ring> = self.ring.borrow().clone();
		self.persist_cluster_layout
			.save_async(&ring.layout)
			.await
			.expect("Cannot save current cluster layout");
		Ok(())
	}

	fn update_local_status(&self) {
		let mut local_status = self.local_status.write().unwrap();

		let ring = self.ring.borrow();
		local_status.cluster_layout_version = ring.layout.version;
		local_status.cluster_layout_staging_hash = ring.layout.staging_hash;

		local_status.update_disk_usage(&self.metadata_dir, &self.data_dir);
	}

	// --- RPC HANDLERS ---

	async fn handle_connect(&self, node: &str) -> Result<SystemRpc, Error> {
		self.connect(node).await?;
		Ok(SystemRpc::Ok)
	}

	fn handle_pull_cluster_layout(&self) -> SystemRpc {
		let ring = self.ring.borrow().clone();
		SystemRpc::AdvertiseClusterLayout(ring.layout.clone())
	}

	fn handle_get_known_nodes(&self) -> SystemRpc {
		let known_nodes = self.get_known_nodes();
		SystemRpc::ReturnKnownNodes(known_nodes)
	}

	async fn handle_advertise_status(
		self: &Arc<Self>,
		from: Uuid,
		info: &NodeStatus,
	) -> Result<SystemRpc, Error> {
		let local_info = self.local_status.read().unwrap();

		if local_info.replication_factor < info.replication_factor {
			error!("Some node have a higher replication factor ({}) than this one ({}). This is not supported and will lead to data corruption. Shutting down for safety.",
				info.replication_factor,
				local_info.replication_factor);
			std::process::exit(1);
		}

		if info.cluster_layout_version > local_info.cluster_layout_version
			|| info.cluster_layout_staging_hash != local_info.cluster_layout_staging_hash
		{
			tokio::spawn(self.clone().pull_cluster_layout(from));
		}

		drop(local_info);

		self.node_status
			.write()
			.unwrap()
			.insert(from, (now_msec(), info.clone()));

		Ok(SystemRpc::Ok)
	}

	async fn handle_advertise_cluster_layout(
		self: &Arc<Self>,
		adv: &ClusterLayout,
	) -> Result<SystemRpc, Error> {
		if adv.replication_factor != self.replication_factor {
			let msg = format!(
				"Received a cluster layout from another node with replication factor {}, which is different from what we have in our configuration ({}). Discarding the cluster layout we received.",
				adv.replication_factor,
				self.replication_factor
			);
			error!("{}", msg);
			return Err(Error::Message(msg));
		}

		let update_ring = self.update_ring.lock().await;
		let mut layout: ClusterLayout = self.ring.borrow().layout.clone();

		let prev_layout_check = layout.check().is_ok();
		if layout.merge(adv) {
			if prev_layout_check && layout.check().is_err() {
				error!("New cluster layout is invalid, discarding.");
				return Err(Error::Message(
					"New cluster layout is invalid, discarding.".into(),
				));
			}

			let ring = Ring::new(layout.clone(), self.replication_factor);
			update_ring.send(Arc::new(ring))?;
			drop(update_ring);

			let self2 = self.clone();
			tokio::spawn(async move {
				if let Err(e) = self2
					.rpc
					.broadcast(
						&self2.system_endpoint,
						SystemRpc::AdvertiseClusterLayout(layout),
						RequestStrategy::with_priority(PRIO_HIGH),
					)
					.await
				{
					warn!("Error while broadcasting new cluster layout: {}", e);
				}
			});

			self.save_cluster_layout().await?;
		}

		Ok(SystemRpc::Ok)
	}

	async fn status_exchange_loop(&self, mut stop_signal: watch::Receiver<bool>) {
		while !*stop_signal.borrow() {
			let restart_at = Instant::now() + STATUS_EXCHANGE_INTERVAL;

			// Update local node status that is exchanged.
			self.update_local_status();

			let local_status: NodeStatus = self.local_status.read().unwrap().clone();
			let _ = self
				.rpc
				.broadcast(
					&self.system_endpoint,
					SystemRpc::AdvertiseStatus(local_status),
					RequestStrategy::with_priority(PRIO_HIGH)
						.with_custom_timeout(STATUS_EXCHANGE_INTERVAL),
				)
				.await;

			select! {
				_ = tokio::time::sleep_until(restart_at.into()) => {},
				_ = stop_signal.changed() => {},
			}
		}
	}

	async fn discovery_loop(self: &Arc<Self>, mut stop_signal: watch::Receiver<bool>) {
		while !*stop_signal.borrow() {
			let not_configured = self.ring.borrow().layout.check().is_err();
			let no_peers = self.peering.get_peer_list().len() < self.replication_factor;
			let expected_n_nodes = self.ring.borrow().layout.num_nodes();
			let bad_peers = self
				.peering
				.get_peer_list()
				.iter()
				.filter(|p| p.is_up())
				.count() != expected_n_nodes;

			if not_configured || no_peers || bad_peers {
				info!("Doing a bootstrap/discovery step (not_configured: {}, no_peers: {}, bad_peers: {})", not_configured, no_peers, bad_peers);

				let mut ping_list = resolve_peers(&self.bootstrap_peers).await;

				// Add peer list from list stored on disk
				if let Ok(peers) = self.persist_peer_list.load_async().await {
					ping_list.extend(peers.0.iter().map(|(id, addr)| ((*id).into(), *addr)))
				}

				// Fetch peer list from Consul
				#[cfg(feature = "consul-discovery")]
				if let Some(c) = &self.consul_discovery {
					match c.get_consul_nodes().await {
						Ok(node_list) => {
							ping_list.extend(node_list);
						}
						Err(e) => {
							warn!("Could not retrieve node list from Consul: {}", e);
						}
					}
				}

				// Fetch peer list from Kubernetes
				#[cfg(feature = "kubernetes-discovery")]
				if let Some(k) = &self.kubernetes_discovery {
					if !k.skip_crd {
						match create_kubernetes_crd().await {
							Ok(()) => (),
							Err(e) => {
								error!("Failed to create kubernetes custom resource: {}", e)
							}
						};
					}

					match get_kubernetes_nodes(k).await {
						Ok(node_list) => {
							ping_list.extend(node_list);
						}
						Err(e) => {
							warn!("Could not retrieve node list from Kubernetes: {}", e);
						}
					}
				}

				for (node_id, node_addr) in ping_list {
					let self2 = self.clone();
					tokio::spawn(async move {
						if let Err(e) = self2.netapp.clone().try_connect(node_addr, node_id).await {
							error!("{}", connect_error_message(node_addr, node_id, e));
						}
					});
				}
			}

			if let Err(e) = self.save_peer_list().await {
				warn!("Could not save peer list to file: {}", e);
			}

			#[cfg(feature = "consul-discovery")]
			tokio::spawn(self.clone().advertise_to_consul());

			#[cfg(feature = "kubernetes-discovery")]
			tokio::spawn(self.clone().advertise_to_kubernetes());

			select! {
				_ = tokio::time::sleep(DISCOVERY_INTERVAL) => {},
				_ = stop_signal.changed() => {},
			}
		}
	}

	async fn save_peer_list(&self) -> Result<(), Error> {
		// Prepare new peer list to save to file
		// It is a vec of tuples (node ID as Uuid, node SocketAddr)
		let mut peer_list = self
			.peering
			.get_peer_list()
			.iter()
			.map(|n| (n.id.into(), n.addr))
			.collect::<Vec<_>>();

		// Before doing it, we read the current peer list file (if it exists)
		// and append it to the list we are about to save,
		// so that no peer ID gets lost in the process.
		if let Ok(mut prev_peer_list) = self.persist_peer_list.load_async().await {
			prev_peer_list
				.0
				.retain(|(id, _ip)| peer_list.iter().all(|(id2, _ip2)| id2 != id));
			peer_list.extend(prev_peer_list.0);
		}

		// Save new peer list to file
		self.persist_peer_list
			.save_async(&PeerList(peer_list))
			.await
	}

	async fn pull_cluster_layout(self: Arc<Self>, peer: Uuid) {
		let resp = self
			.rpc
			.call(
				&self.system_endpoint,
				peer,
				SystemRpc::PullClusterLayout,
				RequestStrategy::with_priority(PRIO_HIGH),
			)
			.await;
		if let Ok(SystemRpc::AdvertiseClusterLayout(layout)) = resp {
			let _: Result<_, _> = self.handle_advertise_cluster_layout(&layout).await;
		}
	}
}

#[async_trait]
impl EndpointHandler<SystemRpc> for System {
	async fn handle(self: &Arc<Self>, msg: &SystemRpc, from: NodeID) -> Result<SystemRpc, Error> {
		match msg {
			SystemRpc::Connect(node) => self.handle_connect(node).await,
			SystemRpc::PullClusterLayout => Ok(self.handle_pull_cluster_layout()),
			SystemRpc::AdvertiseStatus(adv) => self.handle_advertise_status(from.into(), adv).await,
			SystemRpc::AdvertiseClusterLayout(adv) => {
				self.clone().handle_advertise_cluster_layout(adv).await
			}
			SystemRpc::GetKnownNodes => Ok(self.handle_get_known_nodes()),
			m => Err(Error::unexpected_rpc_message(m)),
		}
	}
}

impl NodeStatus {
	fn initial(replication_factor: usize, layout: &ClusterLayout) -> Self {
		NodeStatus {
			hostname: gethostname::gethostname()
				.into_string()
				.unwrap_or_else(|_| "<invalid utf-8>".to_string()),
			replication_factor,
			cluster_layout_version: layout.version,
			cluster_layout_staging_hash: layout.staging_hash,
			meta_disk_avail: None,
			data_disk_avail: None,
		}
	}

	fn unknown() -> Self {
		NodeStatus {
			hostname: "?".to_string(),
			replication_factor: 0,
			cluster_layout_version: 0,
			cluster_layout_staging_hash: Hash::from([0u8; 32]),
			meta_disk_avail: None,
			data_disk_avail: None,
		}
	}

	fn update_disk_usage(&mut self, meta_dir: &Path, data_dir: &DataDirEnum) {
		use nix::sys::statvfs::statvfs;
		let mount_avail = |path: &Path| match statvfs(path) {
			Ok(x) => {
				let avail = x.blocks_available() as u64 * x.fragment_size() as u64;
				let total = x.blocks() as u64 * x.fragment_size() as u64;
				Some((x.filesystem_id(), avail, total))
			}
			Err(_) => None,
		};

		self.meta_disk_avail = mount_avail(meta_dir).map(|(_, a, t)| (a, t));
		self.data_disk_avail = match data_dir {
			DataDirEnum::Single(dir) => mount_avail(dir).map(|(_, a, t)| (a, t)),
			DataDirEnum::Multiple(dirs) => (|| {
				// TODO: more precise calculation that takes into account
				// how data is going to be spread among partitions
				let mut mounts = HashMap::new();
				for dir in dirs.iter() {
					if dir.capacity.is_none() {
						continue;
					}
					match mount_avail(&dir.path) {
						Some((fsid, avail, total)) => {
							mounts.insert(fsid, (avail, total));
						}
						None => return None,
					}
				}
				Some(
					mounts
						.into_iter()
						.fold((0, 0), |(x, y), (_, (a, b))| (x + a, y + b)),
				)
			})(),
		};
	}
}

fn get_default_ip() -> Option<IpAddr> {
	pnet_datalink::interfaces()
		.iter()
		.find(|e| e.is_up() && !e.is_loopback() && !e.ips.is_empty())
		.and_then(|e| e.ips.first())
		.map(|a| a.ip())
}

async fn resolve_peers(peers: &[String]) -> Vec<(NodeID, SocketAddr)> {
	let mut ret = vec![];

	for peer in peers.iter() {
		match parse_and_resolve_peer_addr_async(peer).await {
			Some((pubkey, addrs)) => {
				for ip in addrs {
					ret.push((pubkey, ip));
				}
			}
			None => {
				warn!("Unable to parse and/or resolve peer hostname {}", peer);
			}
		}
	}

	ret
}

fn connect_error_message(
	addr: SocketAddr,
	pubkey: ed25519::PublicKey,
	e: garage_net::error::Error,
) -> String {
	format!("Error establishing RPC connection to remote node: {}@{}.\nThis can happen if the remote node is not reachable on the network, but also if the two nodes are not configured with the same rpc_secret.\n{}", hex::encode(pubkey), addr, e)
}