refactor(telemetry): Standardize compute scheduler metrics

This change updates the OpenDC compute service implementation with
multiple meters that follow the OpenTelemetry conventions.

3 files changed, 48 insertions, 43 deletions

diff --git a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/Portfolio.kt b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/Portfolio.kt
index f7f9336e..3ec424f1 100644
--- a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/Portfolio.kt
+++ b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/Portfolio.kt
@@ -153,11 +153,12 @@ abstract class Portfolio(name: String) : Experiment(name) {
 
         val monitorResults = collectServiceMetrics(clock.millis(), simulator.producers[0])
         logger.debug {
-            "Finish " +
-                "SUBMIT=${monitorResults.instanceCount} " +
-                "FAIL=${monitorResults.failedInstanceCount} " +
-                "QUEUE=${monitorResults.queuedInstanceCount} " +
-                "RUNNING=${monitorResults.activeHostCount}"
+            "Scheduler " +
+                "Success=${monitorResults.attemptsSuccess} " +
+                "Failure=${monitorResults.attemptsFailure} " +
+                "Error=${monitorResults.attemptsError} " +
+                "Pending=${monitorResults.serversPending} " +
+                "Active=${monitorResults.serversActive}"
         }
     }
 }
diff --git a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/export/parquet/ParquetServiceDataWriter.kt b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/export/parquet/ParquetServiceDataWriter.kt
index e1428fe7..29b48878 100644
--- a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/export/parquet/ParquetServiceDataWriter.kt
+++ b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/export/parquet/ParquetServiceDataWriter.kt
@@ -36,13 +36,13 @@ public class ParquetServiceDataWriter(path: File, bufferSize: Int) :
 
     override fun convert(builder: GenericRecordBuilder, data: ServiceData) {
         builder["timestamp"] = data.timestamp
-        builder["host_total_count"] = data.hostCount
-        builder["host_available_count"] = data.activeHostCount
-        builder["instance_total_count"] = data.instanceCount
-        builder["instance_active_count"] = data.runningInstanceCount
-        builder["instance_inactive_count"] = data.finishedInstanceCount
-        builder["instance_waiting_count"] = data.queuedInstanceCount
-        builder["instance_failed_count"] = data.failedInstanceCount
+        builder["hosts_up"] = data.hostsUp
+        builder["hosts_down"] = data.hostsDown
+        builder["servers_pending"] = data.serversPending
+        builder["servers_active"] = data.serversActive
+        builder["attempts_success"] = data.attemptsSuccess
+        builder["attempts_failure"] = data.attemptsFailure
+        builder["attempts_error"] = data.attemptsError
     }
 
     override fun toString(): String = "service-writer"
@@ -53,13 +53,13 @@ public class ParquetServiceDataWriter(path: File, bufferSize: Int) :
             .namespace("org.opendc.telemetry.compute")
             .fields()
             .requiredLong("timestamp")
-            .requiredInt("host_total_count")
-            .requiredInt("host_available_count")
-            .requiredInt("instance_total_count")
-            .requiredInt("instance_active_count")
-            .requiredInt("instance_inactive_count")
-            .requiredInt("instance_waiting_count")
-            .requiredInt("instance_failed_count")
+            .requiredInt("hosts_up")
+            .requiredInt("hosts_down")
+            .requiredInt("servers_pending")
+            .requiredInt("servers_active")
+            .requiredInt("attempts_success")
+            .requiredInt("attempts_failure")
+            .requiredInt("attempts_error")
             .endRecord()
     }
 }
diff --git a/opendc-experiments/opendc-experiments-capelin/src/test/kotlin/org/opendc/experiments/capelin/CapelinIntegrationTest.kt b/opendc-experiments/opendc-experiments-capelin/src/test/kotlin/org/opendc/experiments/capelin/CapelinIntegrationTest.kt
index f4cf3e5e..81405acf 100644
--- a/opendc-experiments/opendc-experiments-capelin/src/test/kotlin/org/opendc/experiments/capelin/CapelinIntegrationTest.kt
+++ b/opendc-experiments/opendc-experiments-capelin/src/test/kotlin/org/opendc/experiments/capelin/CapelinIntegrationTest.kt
@@ -104,19 +104,20 @@ class CapelinIntegrationTest {
 
         val serviceMetrics = collectServiceMetrics(clock.millis(), simulator.producers[0])
         println(
-            "Finish " +
-                "SUBMIT=${serviceMetrics.instanceCount} " +
-                "FAIL=${serviceMetrics.failedInstanceCount} " +
-                "QUEUE=${serviceMetrics.queuedInstanceCount} " +
-                "RUNNING=${serviceMetrics.runningInstanceCount}"
+            "Scheduler " +
+                "Success=${serviceMetrics.attemptsSuccess} " +
+                "Failure=${serviceMetrics.attemptsFailure} " +
+                "Error=${serviceMetrics.attemptsError} " +
+                "Pending=${serviceMetrics.serversPending} " +
+                "Active=${serviceMetrics.serversActive}"
         )
 
         // Note that these values have been verified beforehand
         assertAll(
-            { assertEquals(50, serviceMetrics.instanceCount, "The trace contains 50 VMs") },
-            { assertEquals(0, serviceMetrics.runningInstanceCount, "All VMs should finish after a run") },
-            { assertEquals(0, serviceMetrics.failedInstanceCount, "No VM should not be unscheduled") },
-            { assertEquals(0, serviceMetrics.queuedInstanceCount, "No VM should not be in the queue") },
+            { assertEquals(50, serviceMetrics.attemptsSuccess, "The scheduler should schedule 50 VMs") },
+            { assertEquals(0, serviceMetrics.serversActive, "All VMs should finish after a run") },
+            { assertEquals(0, serviceMetrics.attemptsFailure, "No VM should be unscheduled") },
+            { assertEquals(0, serviceMetrics.serversPending, "No VM should not be in the queue") },
             { assertEquals(220346412191, monitor.totalWork) { "Incorrect requested burst" } },
             { assertEquals(206667852689, monitor.totalGrantedWork) { "Incorrect granted burst" } },
             { assertEquals(1151612221, monitor.totalOvercommittedWork) { "Incorrect overcommitted burst" } },
@@ -152,11 +153,12 @@ class CapelinIntegrationTest {
 
         val serviceMetrics = collectServiceMetrics(clock.millis(), simulator.producers[0])
         println(
-            "Finish " +
-                "SUBMIT=${serviceMetrics.instanceCount} " +
-                "FAIL=${serviceMetrics.failedInstanceCount} " +
-                "QUEUE=${serviceMetrics.queuedInstanceCount} " +
-                "RUNNING=${serviceMetrics.runningInstanceCount}"
+            "Scheduler " +
+                "Success=${serviceMetrics.attemptsSuccess} " +
+                "Failure=${serviceMetrics.attemptsFailure} " +
+                "Error=${serviceMetrics.attemptsError} " +
+                "Pending=${serviceMetrics.serversPending} " +
+                "Active=${serviceMetrics.serversActive}"
         )
 
         // Note that these values have been verified beforehand
@@ -202,11 +204,12 @@ class CapelinIntegrationTest {
 
         val serviceMetrics = collectServiceMetrics(clock.millis(), simulator.producers[0])
         println(
-            "Finish " +
-                "SUBMIT=${serviceMetrics.instanceCount} " +
-                "FAIL=${serviceMetrics.failedInstanceCount} " +
-                "QUEUE=${serviceMetrics.queuedInstanceCount} " +
-                "RUNNING=${serviceMetrics.runningInstanceCount}"
+            "Scheduler " +
+                "Success=${serviceMetrics.attemptsSuccess} " +
+                "Failure=${serviceMetrics.attemptsFailure} " +
+                "Error=${serviceMetrics.attemptsError} " +
+                "Pending=${serviceMetrics.serversPending} " +
+                "Active=${serviceMetrics.serversActive}"
         )
 
         // Note that these values have been verified beforehand
@@ -246,11 +249,12 @@ class CapelinIntegrationTest {
 
         val serviceMetrics = collectServiceMetrics(clock.millis(), simulator.producers[0])
         println(
-            "Finish " +
-                "SUBMIT=${serviceMetrics.instanceCount} " +
-                "FAIL=${serviceMetrics.failedInstanceCount} " +
-                "QUEUE=${serviceMetrics.queuedInstanceCount} " +
-                "RUNNING=${serviceMetrics.runningInstanceCount}"
+            "Scheduler " +
+                "Success=${serviceMetrics.attemptsSuccess} " +
+                "Failure=${serviceMetrics.attemptsFailure} " +
+                "Error=${serviceMetrics.attemptsError} " +
+                "Pending=${serviceMetrics.serversPending} " +
+                "Active=${serviceMetrics.serversActive}"
         )
 
         // Note that these values have been verified beforehand