From e58998e2187886f371de07afc7a059a8c504d50e Mon Sep 17 00:00:00 2001
From: Charlie Voiselle <464492+angrycub@users.noreply.github.com>
Date: Thu, 8 Sep 2022 14:31:36 -0400
Subject: [PATCH] Add client scheduling eligibility to heartbeat (#14483)

---
 .changelog/14483.txt     |  3 +++
 client/client.go         |  8 ++++++++
 nomad/node_endpoint.go   | 12 ++++++++----
 nomad/structs/structs.go |  4 ++++
 4 files changed, 23 insertions(+), 4 deletions(-)
 create mode 100644 .changelog/14483.txt

diff --git a/.changelog/14483.txt b/.changelog/14483.txt
new file mode 100644
index 000000000..07a3e141d
--- /dev/null
+++ b/.changelog/14483.txt
@@ -0,0 +1,3 @@
+```release-note:bug
+metrics: Update client `node_scheduling_eligibility` value with server heartbeats.
+```
diff --git a/client/client.go b/client/client.go
index 3f4042b33..1fa2cc512 100644
--- a/client/client.go
+++ b/client/client.go
@@ -1955,6 +1955,14 @@ func (c *Client) updateNodeStatus() error {
 		}
 	}
 
+	// Check heartbeat response for information about the server-side scheduling
+	// state of this node
+	c.UpdateConfig(func(c *config.Config) {
+		if resp.SchedulingEligibility != "" {
+			c.Node.SchedulingEligibility = resp.SchedulingEligibility
+		}
+	})
+
 	// Update the number of nodes in the cluster so we can adjust our server
 	// rebalance rate.
 	c.servers.SetNumNodes(resp.NumNodes)
diff --git a/nomad/node_endpoint.go b/nomad/node_endpoint.go
index c13ec70cc..689b1082e 100644
--- a/nomad/node_endpoint.go
+++ b/nomad/node_endpoint.go
@@ -199,7 +199,7 @@ func (n *Node) Register(args *structs.NodeRegisterRequest, reply *structs.NodeUp
 
 	n.srv.peerLock.RLock()
 	defer n.srv.peerLock.RUnlock()
-	if err := n.constructNodeServerInfoResponse(snap, reply); err != nil {
+	if err := n.constructNodeServerInfoResponse(args.Node.ID, snap, reply); err != nil {
 		n.logger.Error("failed to populate NodeUpdateResponse", "error", err)
 		return err
 	}
@@ -258,7 +258,7 @@ func equalDevices(n1, n2 *structs.Node) bool {
 }
 
 // updateNodeUpdateResponse assumes the n.srv.peerLock is held for reading.
-func (n *Node) constructNodeServerInfoResponse(snap *state.StateSnapshot, reply *structs.NodeUpdateResponse) error {
+func (n *Node) constructNodeServerInfoResponse(nodeID string, snap *state.StateSnapshot, reply *structs.NodeUpdateResponse) error {
 	reply.LeaderRPCAddr = string(n.srv.raft.Leader())
 
 	// Reply with config information required for future RPC requests
@@ -271,6 +271,10 @@ func (n *Node) constructNodeServerInfoResponse(snap *state.StateSnapshot, reply
 			})
 	}
 
+	// Add ClientStatus information to heartbeat response.
+	node, _ := snap.NodeByID(nil, nodeID)
+	reply.SchedulingEligibility = node.SchedulingEligibility
+
 	// TODO(sean@): Use an indexed node count instead
 	//
 	// Snapshot is used only to iterate over all nodes to create a node
@@ -564,7 +568,7 @@ func (n *Node) UpdateStatus(args *structs.NodeUpdateStatusRequest, reply *struct
 	reply.Index = index
 	n.srv.peerLock.RLock()
 	defer n.srv.peerLock.RUnlock()
-	if err := n.constructNodeServerInfoResponse(snap, reply); err != nil {
+	if err := n.constructNodeServerInfoResponse(node.GetID(), snap, reply); err != nil {
 		n.logger.Error("failed to populate NodeUpdateResponse", "error", err)
 		return err
 	}
@@ -821,7 +825,7 @@ func (n *Node) Evaluate(args *structs.NodeEvaluateRequest, reply *structs.NodeUp
 
 	n.srv.peerLock.RLock()
 	defer n.srv.peerLock.RUnlock()
-	if err := n.constructNodeServerInfoResponse(snap, reply); err != nil {
+	if err := n.constructNodeServerInfoResponse(node.GetID(), snap, reply); err != nil {
 		n.logger.Error("failed to populate NodeUpdateResponse", "error", err)
 		return err
 	}
diff --git a/nomad/structs/structs.go b/nomad/structs/structs.go
index b5e5a25bf..819b378e7 100644
--- a/nomad/structs/structs.go
+++ b/nomad/structs/structs.go
@@ -1361,6 +1361,10 @@ type NodeUpdateResponse struct {
 	// region.
 	Servers []*NodeServerInfo
 
+	// SchedulingEligibility is used to inform clients what the server-side
+	// has for their scheduling status during heartbeats.
+	SchedulingEligibility string
+
 	QueryMeta
 }