yrenat commented on code in PR #6046:
URL: https://github.com/apache/texera/pull/6046#discussion_r3852838879
##########
computing-unit-managing-service/src/main/scala/org/apache/texera/service/resource/ComputingUnitManagingResource.scala:
##########
@@ -61,18 +63,200 @@ import org.apache.texera.service.util.{
KubernetesClient
}
import org.jooq.{DSLContext, EnumType}
+import org.jooq.impl.DSL.max
+import org.slf4j.LoggerFactory
import play.api.libs.json._
import java.sql.Timestamp
import scala.annotation.unused
import scala.jdk.CollectionConverters.CollectionHasAsScala
+import scala.util.control.NonFatal
object ComputingUnitManagingResource {
+ private val logger =
LoggerFactory.getLogger(classOf[ComputingUnitManagingResource])
+
private def context: DSLContext =
SqlServer
.getInstance()
.createDSLContext()
+ private[resource] final class IdleComputingUnitCleanupConfig(
+ val enabled: Boolean,
+ val idleTimeoutMinutes: Long
+ ) {
+ def copy(
+ enabled: Boolean = this.enabled,
+ idleTimeoutMinutes: Long = this.idleTimeoutMinutes
+ ): IdleComputingUnitCleanupConfig =
+ new IdleComputingUnitCleanupConfig(enabled, idleTimeoutMinutes)
+ }
+
+ private[resource] final class IdleComputingUnitCandidate(
+ val unit: WorkflowComputingUnit,
+ val username: Option[String]
+ )
+
+ private[resource] trait KubernetesPodOperations {
+ val podExists: Int => Boolean
+ val deletePod: Int => Unit
+ }
+
+ private[resource] object DefaultKubernetesPodOperations extends
KubernetesPodOperations {
+ private[resource] var podExistsDelegate: Int => Boolean =
KubernetesClient.podExists
+ private[resource] var deletePodDelegate: Int => Unit =
KubernetesClient.deletePod
+
+ override val podExists: Int => Boolean = cuid => podExistsDelegate(cuid)
+ override val deletePod: Int => Unit = cuid => deletePodDelegate(cuid)
+ }
+
+ private[resource] def lastComputingUnitActivityTime(
+ unit: WorkflowComputingUnit,
+ latestUpdateTime: Option[Timestamp],
+ latestStartTime: Option[Timestamp]
+ ): Timestamp =
+ Seq(
+ latestUpdateTime,
+ latestStartTime,
+ Option(unit.getCreationTime)
+ ).flatten.maxBy(_.getTime)
+
+ private[resource] def shouldTerminateIdleComputingUnit(
+ hasActiveExecution: Boolean,
+ lastExecutionTime: Timestamp,
+ cutoff: Timestamp
+ ): Boolean =
+ !hasActiveExecution && lastExecutionTime.before(cutoff)
+
+ def terminateIdleKubernetesComputingUnits():
List[TerminatedComputingUnitInfo] =
+ terminateIdleKubernetesComputingUnits(
+ new IdleComputingUnitCleanupConfig(
+ KubernetesConfig.kubernetesComputingUnitEnabled,
+ KubernetesConfig.computingUnitIdleTimeoutMinutes
+ ),
+ () => new Timestamp(System.currentTimeMillis()),
+ DefaultKubernetesPodOperations
+ )
+
+ private[resource] def terminateIdleKubernetesComputingUnits(
+ cleanupConfig: IdleComputingUnitCleanupConfig,
+ currentTime: () => Timestamp,
+ podOperations: KubernetesPodOperations
+ ): List[TerminatedComputingUnitInfo] = {
+ if (!cleanupConfig.enabled || cleanupConfig.idleTimeoutMinutes <= 0) {
+ return List.empty
+ }
+
+ val now = currentTime()
+ val cutoff = new Timestamp(now.getTime - cleanupConfig.idleTimeoutMinutes
* 60 * 1000)
+
+ idleKubernetesComputingUnitCandidates(cutoff).flatMap(candidate =>
+ terminateIdleKubernetesComputingUnitCandidate(candidate, now,
podOperations)
+ )
+ }
+
+ private[resource] def idleKubernetesComputingUnitCandidates(
+ cutoff: Timestamp
+ ): List[IdleComputingUnitCandidate] = {
+ val activeStatuses = Seq(Short.box(0), Short.box(1), Short.box(2))
+
+ withTransaction(context) { ctx =>
+ val userDao = new UserDao(ctx.configuration())
+ ctx
+ .selectFrom(WORKFLOW_COMPUTING_UNIT)
+ .where(
+ WORKFLOW_COMPUTING_UNIT.TYPE
+ .eq(WorkflowComputingUnitTypeEnum.kubernetes)
+ .and(WORKFLOW_COMPUTING_UNIT.TERMINATE_TIME.isNull)
+ )
+ .fetchInto(classOf[WorkflowComputingUnit])
+ .asScala
+ .flatMap { unit =>
+ val cuid = unit.getCuid
+ val hasActiveExecution = ctx.fetchExists(
+ ctx
+ .selectOne()
+ .from(WORKFLOW_EXECUTIONS)
+ .where(
+ WORKFLOW_EXECUTIONS.CUID
+ .eq(cuid)
+ .and(WORKFLOW_EXECUTIONS.STATUS.in(activeStatuses: _*))
+ )
+ )
+ val latestUpdateTime = ctx
+ .select(max(WORKFLOW_EXECUTIONS.LAST_UPDATE_TIME))
+ .from(WORKFLOW_EXECUTIONS)
+ .where(WORKFLOW_EXECUTIONS.CUID.eq(cuid))
+ .fetchOne(0, classOf[Timestamp])
+ val latestStartTime = ctx
+ .select(max(WORKFLOW_EXECUTIONS.STARTING_TIME))
+ .from(WORKFLOW_EXECUTIONS)
Review Comment:
Addressed. The scan is now one `SELECT … GROUP BY cuid, user.name` with the
two `LEFT JOIN`s, and the `fetchExists`, both `max()es` and the `fetchOneByUid`
are gone. I keep `shouldTerminateIdleComputingUnit` in Scala rather than a
`HAVING`. I also added `idx_workflow_executions_cuid` in both `38.sql` and
`texera_ddl.sql`, so existing and fresh deployments both get it.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
