kvserver: improve observability with decommission nudger

Previously, we added the decommissioning nudger which nudges the leaseholder
replica of decommissioning ranges to enqueue themselves into the replicate queue
for decommissioning. However, we are still observing extended decommission stall
with the nudger enabled. Observability was limited, and we could not easily tell
whether replicas were successfully enqueued or processed.

This commit improves observability by adding four metrics to track the enqueue
and processing results of the decommissioning nudger:
ranges.decommissioning.nudger.{enqueue,process}.{success,failure}.

Author: wenyihu6

docs/generated/metrics/metrics.html

@@ -598,7 +598,11 @@
 <tr><td>STORAGE</td><td>ranges</td><td>Number of ranges</td><td>Ranges</td><td>GAUGE</td><td>COUNT</td><td>AVG</td><td>NONE</td></tr>
 <tr><td>STORAGE</td><td>ranges.decommissioning</td><td>Number of ranges with at lease one replica on a decommissioning node</td><td>Ranges</td><td>GAUGE</td><td>COUNT</td><td>AVG</td><td>NONE</td></tr>
 <tr><td>STORAGE</td><td>ranges.decommissioning.nudger.enqueue</td><td>Number of enqueued enqueues of a range for decommissioning by the decommissioning nudger. Note: This metric tracks when the nudger attempts to enqueue, but the replica might not end up being enqueued by the priority queue due to various filtering or failure conditions.</td><td>Ranges</td><td>COUNTER</td><td>COUNT</td><td>AVG</td><td>NON_NEGATIVE_DERIVATIVE</td></tr>
-<tr><td>STORAGE</td><td>ranges.decommissioning.nudger.not_leaseholder_or_invalid_lease</td><td>Number of enqueues of a range for decommissioning by the decommissioning nudger that were not the leaseholder or had an invalid lease</td><td>Ranges</td><td>COUNTER</td><td>COUNT</td><td>AVG</td><td>NON_NEGATIVE_DERIVATIVE</td></tr>
+<tr><td>STORAGE</td><td>ranges.decommissioning.nudger.enqueue.failure</td><td>Number of ranges that failed to enqueue at the replicate queue</td><td>Ranges</td><td>COUNTER</td><td>COUNT</td><td>AVG</td><td>NON_NEGATIVE_DERIVATIVE</td></tr>
+<tr><td>STORAGE</td><td>ranges.decommissioning.nudger.enqueue.success</td><td>Number of ranges that were successfully enqueued by the decommisioning nudger</td><td>Ranges</td><td>COUNTER</td><td>COUNT</td><td>AVG</td><td>NON_NEGATIVE_DERIVATIVE</td></tr>
+<tr><td>STORAGE</td><td>ranges.decommissioning.nudger.not_leaseholder_or_invalid_lease</td><td>Number of ranges that were not the leaseholder or had an invalid lease at the decommissioning nudger</td><td>Ranges</td><td>COUNTER</td><td>COUNT</td><td>AVG</td><td>NON_NEGATIVE_DERIVATIVE</td></tr>
+<tr><td>STORAGE</td><td>ranges.decommissioning.nudger.process.failure</td><td>Number of ranges enqueued by the decommissioning nudger that failed to process by the replicate queue</td><td>Ranges</td><td>COUNTER</td><td>COUNT</td><td>AVG</td><td>NON_NEGATIVE_DERIVATIVE</td></tr>
+<tr><td>STORAGE</td><td>ranges.decommissioning.nudger.process.success</td><td>Number of ranges enqueued by the decommissioning nudger that were successfully processed by the replicate queue</td><td>Ranges</td><td>COUNTER</td><td>COUNT</td><td>AVG</td><td>NON_NEGATIVE_DERIVATIVE</td></tr>
 <tr><td>STORAGE</td><td>ranges.overreplicated</td><td>Number of ranges with more live replicas than the replication target</td><td>Ranges</td><td>GAUGE</td><td>COUNT</td><td>AVG</td><td>NONE</td></tr>
 <tr><td>STORAGE</td><td>ranges.unavailable</td><td>Number of ranges with fewer live replicas than needed for quorum</td><td>Ranges</td><td>GAUGE</td><td>COUNT</td><td>AVG</td><td>NONE</td></tr>
 <tr><td>STORAGE</td><td>ranges.underreplicated</td><td>Number of ranges with fewer live replicas than the replication target</td><td>Ranges</td><td>GAUGE</td><td>COUNT</td><td>AVG</td><td>NONE</td></tr>

pkg/kv/kvserver/metrics.go

@@ -149,9 +149,33 @@ var (
 		Measurement: "Ranges",
 		Unit:        metric.Unit_COUNT,
 	}
+	metaDecommissioningNudgerEnqueueSuccess = metric.Metadata{
+		Name:        "ranges.decommissioning.nudger.enqueue.success",
+		Help:        "Number of ranges that were successfully enqueued by the decommisioning nudger",
+		Measurement: "Ranges",
+		Unit:        metric.Unit_COUNT,
+	}
+	metaDecommissioningNudgerEnqueueFailure = metric.Metadata{
+		Name:        "ranges.decommissioning.nudger.enqueue.failure",
+		Help:        "Number of ranges that failed to enqueue at the replicate queue",
+		Measurement: "Ranges",
+		Unit:        metric.Unit_COUNT,
+	}
+	metaDecommissioningNudgerProcessSuccess = metric.Metadata{
+		Name:        "ranges.decommissioning.nudger.process.success",
+		Help:        "Number of ranges enqueued by the decommissioning nudger that were successfully processed by the replicate queue",
+		Measurement: "Ranges",
+		Unit:        metric.Unit_COUNT,
+	}
+	metaDecommissioningNudgerProcessFailure = metric.Metadata{
+		Name:        "ranges.decommissioning.nudger.process.failure",
+		Help:        "Number of ranges enqueued by the decommissioning nudger that failed to process by the replicate queue",
+		Measurement: "Ranges",
+		Unit:        metric.Unit_COUNT,
+	}
 	metaDecommissioningNudgerNotLeaseholderOrInvalidLease = metric.Metadata{
 		Name:        "ranges.decommissioning.nudger.not_leaseholder_or_invalid_lease",
-		Help:        "Number of enqueues of a range for decommissioning by the decommissioning nudger that were not the leaseholder or had an invalid lease",
+		Help:        "Number of ranges that were not the leaseholder or had an invalid lease at the decommissioning nudger",
 		Measurement: "Ranges",
 		Unit:        metric.Unit_COUNT,
 	}
@@ -2653,6 +2677,10 @@ type StoreMetrics struct {
 
 	// Decommissioning nudger metrics.
 	DecommissioningNudgerEnqueue                      *metric.Counter
+	DecommissioningNudgerEnqueueSuccess               *metric.Counter
+	DecommissioningNudgerEnqueueFailure               *metric.Counter
+	DecommissioningNudgerProcessSuccess               *metric.Counter
+	DecommissioningNudgerProcessFailure               *metric.Counter
 	DecommissioningNudgerNotLeaseholderOrInvalidLease *metric.Counter
 
 	// Lease request metrics for successful and failed lease requests. These
@@ -3362,6 +3390,10 @@ func newStoreMetrics(histogramWindow time.Duration) *StoreMetrics {
 
 		// Decommissioning nuder metrics.
 		DecommissioningNudgerEnqueue:                      metric.NewCounter(metaDecommissioningNudgerEnqueue),
+		DecommissioningNudgerEnqueueSuccess:               metric.NewCounter(metaDecommissioningNudgerEnqueueSuccess),
+		DecommissioningNudgerEnqueueFailure:               metric.NewCounter(metaDecommissioningNudgerEnqueueFailure),
+		DecommissioningNudgerProcessSuccess:               metric.NewCounter(metaDecommissioningNudgerProcessSuccess),
+		DecommissioningNudgerProcessFailure:               metric.NewCounter(metaDecommissioningNudgerProcessFailure),
 		DecommissioningNudgerNotLeaseholderOrInvalidLease: metric.NewCounter(metaDecommissioningNudgerNotLeaseholderOrInvalidLease),
 
 		// Lease request metrics.

pkg/kv/kvserver/replica.go

@@ -2618,18 +2618,22 @@ func (r *Replica) maybeEnqueueProblemRange(
 				if err != nil {
 					log.KvDistribution.VInfof(ctx, level,
 						"decommissioning nudger failed to enqueue range %v due to %v", r.Desc(), err)
+					r.store.metrics.DecommissioningNudgerEnqueueFailure.Inc(1)
 				} else {
 					log.KvDistribution.VInfof(ctx, level,
 						"decommissioning nudger successfully enqueued range %v at index %d", r.Desc(), indexOnHeap)
+					r.store.metrics.DecommissioningNudgerEnqueueSuccess.Inc(1)
 				}
 			},
 			onProcessResult: func(err error) {
 				if err != nil {
 					log.KvDistribution.VInfof(ctx, level,
 						"decommissioning nudger failed to process range %v due to %v", r.Desc(), err)
+					r.store.metrics.DecommissioningNudgerProcessFailure.Inc(1)
 				} else {
 					log.KvDistribution.VInfof(ctx, level,
 						"decommissioning nudger successfully processed replica %s", r.Desc())
+					r.store.metrics.DecommissioningNudgerProcessSuccess.Inc(1)
 				}
 			},
 		})