Bring back PeerConnected and PeerDisconnected

Indicate connectivity to peers by also switching from 'NodeId' to 'Host'. This will make sure that the reported connection/disconnection is associated with --peer command line arguments.
cardano-scaling · Feb 22, 2025 · 3e52f27 · 3e52f27
1 parent e1a6272
commit 3e52f27
Show file tree

Hide file tree

Showing 16 changed files with 197 additions and 68 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -11,7 +11,9 @@ changes.
 ## [0.21.0] - UNRELEASED
 
 - **BREAKING** Switch to using `etcd` internally to establish a reliable L2 network
-  - Drops `PeerConnected` and `PeerDisconnected` messages in favor of `NetworkConnected` and `NetworkDisconnected` equivalents. These now indicate overall connectivity to the L2 network (which corresponds to the `etcd` cluster).
+  - New run-time dependency onto `etcd` binary
+  - Adds `NetworkConnected` and `NetworkDisconnected` outputs which are most indicative of whether the L2 network is up or not.
+  - Change `PeerConnected` and `PeerDisconnected` to indicate connectivity to `--peer` items and not the remote `node-id`.
   - Log outputs related to the network components changed significantly.
   - Persisted state (write ahead logs) of the network components changed significantly.
 
@@ -20,7 +22,7 @@ changes.
   - An example was if the datum would contain CBOR instead of just hex encoded bytes.
 
 - **BREAKING** Enable multi-party, networked "offline" heads by providing an `--offline-head-seed` option to `hydra-node`.
-  - Drop `hydra-nodde offline` as a sub-command. Use `--offline-head-seed` and `--initial-utxo` options to switch to offline mode.
+  - Drop `hydra-node offline` as a sub-command. Use `--offline-head-seed` and `--initial-utxo` options to switch to offline mode.
 
 - Stream historical data from disk in the hydra-node API server.
 

diff --git a/hydra-node/golden/ServerOutput/NetworkConnected.json b/hydra-node/golden/ServerOutput/NetworkConnected.json
@@ -0,0 +1,8 @@
+{
+    "samples": [
+        {
+            "tag": "NetworkConnected"
+        }
+    ],
+    "seed": -2100000504
+}
diff --git a/hydra-node/golden/ServerOutput/NetworkDisconnected.json b/hydra-node/golden/ServerOutput/NetworkDisconnected.json
@@ -0,0 +1,8 @@
+{
+    "samples": [
+        {
+            "tag": "NetworkDisconnected"
+        }
+    ],
+    "seed": -1678395060
+}
diff --git a/hydra-node/golden/ServerOutput/PeerConnected.json b/hydra-node/golden/ServerOutput/PeerConnected.json
@@ -1,9 +1,12 @@
 {
     "samples": [
         {
-            "peer": "qitfwqawhsprnrebktoraqk",
+            "peer": {
+                "hostname": "0.0.122.110",
+                "port": 1964
+            },
             "tag": "PeerConnected"
         }
     ],
-    "seed": 662696084
+    "seed": 1214056638
 }
diff --git a/hydra-node/golden/ServerOutput/PeerDisconnected.json b/hydra-node/golden/ServerOutput/PeerDisconnected.json
@@ -1,9 +1,12 @@
 {
     "samples": [
         {
-            "peer": "zocryeobfgljgxe",
+            "peer": {
+                "hostname": "0.0.47.58",
+                "port": 17398
+            },
             "tag": "PeerDisconnected"
         }
     ],
-    "seed": -1791938521
+    "seed": 966094347
 }
diff --git a/hydra-node/json-schemas/api.yaml b/hydra-node/json-schemas/api.yaml
@@ -73,6 +73,8 @@ channels:
       message:
         oneOf:
           - $ref: "api.yaml#/components/messages/Greetings"
+          - $ref: "api.yaml#/components/messages/PeerConnected"
+          - $ref: "api.yaml#/components/messages/PeerDisconnected"
           - $ref: "api.yaml#/components/messages/NetworkConnected"
           - $ref: "api.yaml#/components/messages/NetworkDisconnected"
           - $ref: "api.yaml#/components/messages/PeerHandshakeFailure"
@@ -458,6 +460,9 @@ components:
         $ref: "api.yaml#/components/schemas/Greetings"
 
     NetworkConnected:
+      title: NetworkConnected
+      description: |
+        Node connected to the L2 Hydra network.
       payload:
         $ref: "api.yaml#/components/schemas/NetworkConnected"
 
@@ -468,6 +473,21 @@ components:
       payload:
         $ref: "api.yaml#/components/schemas/NetworkDisconnected"
 
+    PeerConnected:
+      title: PeerConnected
+      description: |
+        A peer appeared alive on the L2 Hydra network.
+      payload:
+        $ref: "api.yaml#/components/schemas/PeerConnected"
+
+    PeerDisconnected:
+      title: PeerDisconnected
+      description: |
+        A peer disappeared (has not been seen active in a while) from the L2
+        Hydra network.
+      payload:
+        $ref: "api.yaml#/components/schemas/PeerDisconnected"
+
     PeerHandshakeFailure:
       title: PeerHandshakeFailure
       description: |
@@ -707,6 +727,8 @@ components:
         - $ref: "api.yaml#/components/schemas/Greetings"
         - $ref: "api.yaml#/components/schemas/NetworkConnected"
         - $ref: "api.yaml#/components/schemas/NetworkDisconnected"
+        - $ref: "api.yaml#/components/schemas/PeerConnected"
+        - $ref: "api.yaml#/components/schemas/PeerDisconnected"
         - $ref: "api.yaml#/components/schemas/PeerHandshakeFailure"
         - $ref: "api.yaml#/components/schemas/HeadIsInitializing"
         - $ref: "api.yaml#/components/schemas/Committed"
@@ -761,9 +783,6 @@ components:
           $ref: "api.yaml#/components/schemas/HydraNodeVersion"
 
     NetworkConnected:
-      title: NetworkConnected
-      description: |
-        Node connected to the L2 Hydra network.
       type: object
       required:
         - tag
@@ -787,12 +806,49 @@ components:
       properties:
         tag:
           type: string
-          enum: ["NetworkConnected"]
+          enum: ["NetworkDisconnected"]
+        seq:
+          $ref: "api.yaml#/components/schemas/SequenceNumber"
+        timestamp:
+          $ref: "api.yaml#/components/schemas/UTCTime"
+
+    PeerConnected:
+      type: object
+      required:
+        - tag
+        - peer
+        - seq
+        - timestamp
+      properties:
+        tag:
+          type: string
+          enum: ["PeerConnected"]
+        peer:
+          $ref: "api.yaml#/components/schemas/Host"
         seq:
           $ref: "api.yaml#/components/schemas/SequenceNumber"
         timestamp:
           $ref: "api.yaml#/components/schemas/UTCTime"
 
+    PeerDisconnected:
+      type: object
+      required:
+        - tag
+        - peer
+        - seq
+        - timestamp
+      properties:
+        tag:
+          type: string
+          enum: ["PeerDisconnected"]
+        peer:
+          $ref: "api.yaml#/components/schemas/Host"
+        seq:
+          $ref: "api.yaml#/components/schemas/SequenceNumber"
+        timestamp:
+          $ref: "api.yaml#/components/schemas/UTCTime"
+
+    # TODO: Update to indicate network-level incompatibility
     PeerHandshakeFailure:
       type: object
       required:
@@ -808,7 +864,7 @@ components:
           enum: ["PeerHandshakeFailure"]
         remoteHost:
           type: object
-          $ref: "api.yaml#/components/schemas/IP"
+          $ref: "api.yaml#/components/schemas/Host"
         ourVersion:
           type: integer
           minimum: 0
@@ -2640,6 +2696,23 @@ components:
     # NOTE: We don't want anyone to depend on this!
     ChainState: {}
 
+    Host:
+      description: |
+        Describes a host we can connect to using TCP.
+      type: object
+      required:
+        - hostname
+        - port
+      properties:
+        hostname:
+          type: string
+        port:
+          type: integer
+          minimum: 0
+          maximum: 65535
+          example: 5000
+
+
     IP:
       type: object
       oneOf:

diff --git a/hydra-node/src/Hydra/API/ServerOutput.hs b/hydra-node/src/Hydra/API/ServerOutput.hs
@@ -77,6 +77,8 @@ instance ArbitraryIsTx tx => Arbitrary (DecommitInvalidReason tx) where
 data ServerOutput tx
   = NetworkConnected
   | NetworkDisconnected
+  | PeerConnected {peer :: Host}
+  | PeerDisconnected {peer :: Host}
   | PeerHandshakeFailure
       { remoteHost :: Host
       , ourVersion :: Natural
@@ -168,6 +170,8 @@ instance (ArbitraryIsTx tx, IsChainState tx) => Arbitrary (ServerOutput tx) wher
   -- Overlapping instances with 'UTxOType tx' even though for a fixed `tx`, there
   -- should be only one 'UTxOType tx'
   shrink = \case
+    PeerConnected p -> PeerConnected <$> shrink p
+    PeerDisconnected p -> PeerDisconnected <$> shrink p
     NetworkConnected -> pure NetworkConnected
     NetworkDisconnected -> pure NetworkDisconnected
     PeerHandshakeFailure rh ov tv -> PeerHandshakeFailure <$> shrink rh <*> shrink ov <*> shrink tv
@@ -235,6 +239,8 @@ prepareServerOutput ::
   LBS.ByteString
 prepareServerOutput ServerOutputConfig{utxoInSnapshot} response =
   case output response of
+    PeerConnected{} -> encodedResponse
+    PeerDisconnected{} -> encodedResponse
     NetworkConnected{} -> encodedResponse
     NetworkDisconnected{} -> encodedResponse
     PeerHandshakeFailure{} -> encodedResponse

diff --git a/hydra-node/src/Hydra/HeadLogic.hs b/hydra-node/src/Hydra/HeadLogic.hs
@@ -103,11 +103,14 @@ defaultTTL = 5
 
 onConnectionEvent :: Connectivity -> Outcome tx
 onConnectionEvent = \case
-  -- FIXME: bring back PeerConnected and PeerDisconnected
-  NetworkConnected{} ->
+  NetworkConnected ->
     causes [ClientEffect ServerOutput.NetworkConnected]
-  NetworkDisconnected{} ->
+  NetworkDisconnected ->
     causes [ClientEffect ServerOutput.NetworkDisconnected]
+  PeerConnected{peer} ->
+    causes [ClientEffect ServerOutput.PeerConnected{peer}]
+  PeerDisconnected{peer} ->
+    causes [ClientEffect ServerOutput.PeerDisconnected{peer}]
   HandshakeFailure{remoteHost, ourVersion, theirVersions} ->
     causes
       [ ClientEffect

diff --git a/hydra-node/src/Hydra/Network.hs b/hydra-node/src/Hydra/Network.hs
@@ -96,6 +96,25 @@ deriving anyclass instance FromJSON IP
 
 -- ** PortNumber (Orphans)
 
+readPort :: MonadFail m => String -> m PortNumber
+readPort s =
+  case readMaybe s of
+    Nothing -> fail "cannot read port"
+    Just n
+      | n >= minPort && n <= maxPort -> pure $ fromInteger n
+      | otherwise ->
+          fail $
+            "readPort: "
+              <> show n
+              <> " not within valid port range: ("
+              <> show minPort
+              <> ", "
+              <> show maxPort
+              <> ")"
+ where
+  maxPort = fromIntegral (maxBound :: Word16)
+  minPort = fromIntegral (minBound :: Word16)
+
 instance ToJSON PortNumber where
   toJSON = toJSON . toInteger
 
@@ -105,15 +124,15 @@ instance FromJSON PortNumber where
 instance Arbitrary PortNumber where
   arbitrary = fromIntegral @Word16 <$> arbitrary
 
+-- ** NodeId
+
 newtype NodeId = NodeId {nodeId :: Text}
   deriving newtype (Eq, Show, IsString, ToString, Read, Ord, ToJSON, FromJSON)
 
 instance Arbitrary NodeId where
   arbitrary =
     NodeId . pack <$> suchThat (listOf (elements ['a' .. 'z'])) (not . null)
 
--- return $ NodeId $ pack c
-
 instance FromCBOR NodeId where
   fromCBOR = NodeId <$> fromCBOR
 
@@ -163,34 +182,15 @@ readHost s =
     (h, ':' : p) -> Host (pack h) <$> readPort p
     _ -> fail $ "readHost: missing : in " <> s
 
-readPort :: MonadFail m => String -> m PortNumber
-readPort s =
-  case readMaybe s of
-    Nothing -> fail "cannot read port"
-    Just n
-      | n >= minPort && n <= maxPort -> pure $ fromInteger n
-      | otherwise ->
-          fail $
-            "readPort: "
-              <> show n
-              <> " not within valid port range: ("
-              <> show minPort
-              <> ", "
-              <> show maxPort
-              <> ")"
- where
-  maxPort = fromIntegral (maxBound :: Word16)
-  minPort = fromIntegral (minBound :: Word16)
-
 -- ** Connectivity & versions
 
 -- TODO: improve these types
 
 data Connectivity
   = -- | Individual peer appeared alive on network.
-    Connected {peer :: Host}
+    PeerConnected {peer :: Host}
   | -- | Individual peer disappeared from network (has not been seen active in a while).
-    Disconnected {peer :: Host}
+    PeerDisconnected {peer :: Host}
   | -- | Connected to Hydra network.
     NetworkConnected
   | -- | Disconnected from Hydra network.

diff --git a/hydra-node/src/Hydra/Network/Etcd.hs b/hydra-node/src/Hydra/Network/Etcd.hs
@@ -367,8 +367,8 @@ pollConnectivity conn localHost NetworkCallback{onConnectivity} = do
           alive <- getAlive
           let othersAlive = alive \\ [localHost]
           seenAlive <- atomically $ swapTVar seenAliveVar othersAlive
-          forM_ (othersAlive \\ seenAlive) $ onConnectivity . Connected
-          forM_ (seenAlive \\ othersAlive) $ onConnectivity . Disconnected
+          forM_ (othersAlive \\ seenAlive) $ onConnectivity . PeerConnected
+          forM_ (seenAlive \\ othersAlive) $ onConnectivity . PeerDisconnected
           threadDelay 1
  where
   ttl = 3

diff --git a/hydra-node/src/Hydra/Network/Heartbeat.hs b/hydra-node/src/Hydra/Network/Heartbeat.hs
@@ -122,7 +122,7 @@ updateStateFromIncomingMessages heartbeatState callback =
   notifyAlive peer = do
     now <- getMonotonicTime
     aliveSet <- alive <$> readTVarIO heartbeatState
-    unless (peer `Map.member` aliveSet) $ onConnectivity (Connected peer)
+    unless (peer `Map.member` aliveSet) $ onConnectivity (PeerConnected peer)
     atomically $
       modifyTVar' heartbeatState $ \s ->
         s
@@ -177,7 +177,7 @@ checkRemoteParties heartbeatState onConnectivity =
     threadDelay (heartbeatDelay * 2)
     now <- getMonotonicTime
     updateSuspected heartbeatState now
-      >>= mapM_ (onConnectivity . Disconnected)
+      >>= mapM_ (onConnectivity . PeerDisconnected)
 
 updateSuspected :: MonadSTM m => TVar m HeartbeatState -> Time -> m (Set Host)
 updateSuspected heartbeatState now =