m/n/core: implement node heartbeats
This change introduces cluster member node health monitoring by
implementing a bidirectional RPC stream the nodes will periodically
send their heartbeat updates through. Management.GetNodes call was
modified to include the new node health information.
Relevant data available through the management API is non-persistent,
and stored within current Curator leader's local state. As such, it
will become briefly unavailable in an event of leader re-election. The
information returned, however, is guaranteed to be correct.
Change-Id: I916ac48f496941a7decc09d672ecf72a914b0d88
Reviewed-on: https://review.monogon.dev/c/monogon/+/694
Reviewed-by: Sergiusz Bazanski <serge@monogon.tech>
Tested-by: Jenkins CI
diff --git a/metropolis/node/core/roleserve/roleserve.go b/metropolis/node/core/roleserve/roleserve.go
index 667564d..82e7cca 100644
--- a/metropolis/node/core/roleserve/roleserve.go
+++ b/metropolis/node/core/roleserve/roleserve.go
@@ -75,6 +75,7 @@
controlPlane *workerControlPlane
statusPush *workerStatusPush
+ heartbeat *workerHeartbeat
kubernetes *workerKubernetes
rolefetch *workerRoleFetch
}
@@ -99,6 +100,12 @@
clusterMembership: &s.ClusterMembership,
}
+ s.heartbeat = &workerHeartbeat{
+ network: s.Network,
+
+ clusterMembership: &s.ClusterMembership,
+ }
+
s.kubernetes = &workerKubernetes{
network: s.Network,
storageRoot: s.StorageRoot,
@@ -153,6 +160,7 @@
supervisor.Run(ctx, "controlplane", s.controlPlane.run)
supervisor.Run(ctx, "kubernetes", s.kubernetes.run)
supervisor.Run(ctx, "statuspush", s.statusPush.run)
+ supervisor.Run(ctx, "heartbeat", s.heartbeat.run)
supervisor.Run(ctx, "rolefetch", s.rolefetch.run)
supervisor.Signal(ctx, supervisor.SignalHealthy)