chore(node): improve error on failed send + track dysf

sn_node/src/comm/peer_session.rs

@@ -90,7 +90,7 @@ impl PeerSession {
                     "Failed sending {msg_id:?} to {:?} {err:?}",
                     self.link.peer()
                 );
-                Error::CmdSendError
+                Error::AdultCmdSendError(*self.link.peer())
             })
     }
 

sn_node/src/node/data/records.rs

@@ -144,6 +144,9 @@ impl MyNode {
                 }
                 Err(error) => {
                     error!("{msg_id:?} Error when replicating to adult {peer:?}: {error:?}");
+                    if let Error::AdultCmdSendError(peer) = error {
+                        context.log_node_issue(peer.name(), IssueType::Communication);
+                    }
                     last_error = Some(error);
                 }
             }

sn_node/src/node/error.rs

@@ -158,6 +158,9 @@ pub enum Error {
     /// Error Sending Cmd in to node for processing
     #[error("Error Sending Cmd in to node for processing.")]
     CmdSendError,
+    /// Error Sending Cmd in to node for processing
+    #[error("Error Sending Cmd to adult {0:?} for processing.")]
+    AdultCmdSendError(Peer),
     /// Network Knowledge error.
     #[error("Network data error:: {0}")]
     NetworkKnowledge(#[from] sn_interface::network_knowledge::Error),

sn_node/src/node/mod.rs

@@ -146,6 +146,8 @@ mod core {
         #[debug(skip)]
         pub(crate) comm: Comm,
         pub(crate) joins_allowed: bool,
+        #[debug(skip)]
+        pub(crate) dysfunction_cmds_sender: mpsc::Sender<DysCmds>,
     }
 
     impl NodeContext {
@@ -158,6 +160,21 @@ mod core {
                 .section_auth_by_name(name)
                 .map_err(Error::from)
         }
+
+        /// Log an issue in dysfunction
+        /// Spawns a process to send this incase the channel may be full, we don't hold up
+        /// processing around this (as this can be called during dkg eg)
+        pub(crate) fn log_node_issue(&self, name: XorName, issue: IssueType) {
+            trace!("Logging issue {issue:?} in dysfunction for {name}");
+            let dysf_sender = self.dysfunction_cmds_sender.clone();
+            // TODO: do we need to kill the node if we fail tracking dysf?
+            let _handle = tokio::spawn(async move {
+                if let Err(error) = dysf_sender.send(DysCmds::TrackIssue(name, issue)).await {
+                    // Log the issue, and error. We need to be wary of actually hitting this.
+                    warn!("Could not send DysCmds through dysfunctional_cmds_tx: {error}");
+                }
+            });
+        }
     }
 
     impl MyNode {
@@ -176,6 +193,7 @@ mod core {
                 comm: self.comm.clone(),
                 joins_allowed: self.joins_allowed,
                 data_storage: self.data_storage.clone(),
+                dysfunction_cmds_sender: self.dysfunction_cmds_sender.clone(),
             }
         }