Skip to content

[AMORO-4198][optimizer] Support graceful shutdown for in-progress tasks #4197

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
czy006 merged 6 commits into from
Jul 6, 2026
Merged

[AMORO-4198][optimizer] Support graceful shutdown for in-progress tasks #4197

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
2a6df8f
[AMORO][optimizer] Support graceful shutdown for in-progress tasks
j1wonpark Apr 30, 2026
cd50f29
Merge branch 'master' into optimizer-graceful-shutdown
j1wonpark Jun 30, 2026
99367db
[AMORO][optimizer] Wire shutdown-timeout-ms group property to -st sta…
j1wonpark Jul 4, 2026
21d75de
[AMORO][optimizer] Add graceful drain to Flink optimizer on job cance…
j1wonpark Jul 4, 2026
b4e5ae3
[AMORO][optimizer] Harden the shutdown drain so finished task results…
j1wonpark Jul 4, 2026
e34d2bc
[AMORO][optimizer] Keep the optimizer registration alive during the F…
j1wonpark Jul 4, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions amoro-ams/src/main/java/org/apache/amoro/server/manager/AbstractOptimizerContainer.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -92,6 +92,11 @@ protected String buildOptimizerStartupArgsString(Resource resource) {
.append(" -hb ")
.append(resource.getProperties().get(OptimizerProperties.OPTIMIZER_HEART_BEAT_INTERVAL));
}
if (resource.getProperties().containsKey(OptimizerProperties.OPTIMIZER_SHUTDOWN_TIMEOUT_MS)) {
stringBuilder
.append(" -st ")
.append(resource.getProperties().get(OptimizerProperties.OPTIMIZER_SHUTDOWN_TIMEOUT_MS));
}
if (PropertyUtil.propertyAsBoolean(
resource.getProperties(),
OptimizerProperties.OPTIMIZER_EXTEND_DISK_STORAGE,
Expand Down
46 changes: 40 additions & 6 deletions amoro-ams/src/main/java/org/apache/amoro/server/manager/KubernetesOptimizerContainer.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,10 +34,12 @@
import io.fabric8.kubernetes.client.ConfigBuilder;
import io.fabric8.kubernetes.client.KubernetesClient;
import io.fabric8.kubernetes.client.KubernetesClientBuilder;
import org.apache.amoro.OptimizerProperties;
import org.apache.amoro.resource.Resource;
import org.apache.amoro.shade.guava32.com.google.common.base.Preconditions;
import org.apache.amoro.shade.guava32.com.google.common.collect.ImmutableMap;
import org.apache.amoro.shade.guava32.com.google.common.collect.Maps;
import org.apache.amoro.utils.PropertyUtil;
import org.apache.commons.io.IOUtils;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
Expand Down Expand Up @@ -75,6 +77,13 @@ public class KubernetesOptimizerContainer extends AbstractOptimizerContainer {

private static final String EXTRA_PROPERTY_PREFIX = "extra.";

/**
* Buffer added on top of optimizer's shutdown-timeout when deriving K8s
* terminationGracePeriodSeconds, to leave room for thread join, best-effort task result
* reporting, and log flush before SIGKILL.
*/
static final long TERMINATION_GRACE_BUFFER_SECONDS = 30L;

private static final Map<String, String> EXTRA_PROPERTY_DEFAULTS = new HashMap<>();

static {
Expand All @@ -86,6 +95,16 @@ private String getExtraProperty(Map<String, String> properties, String key) {
EXTRA_PROPERTY_PREFIX + key, EXTRA_PROPERTY_DEFAULTS.getOrDefault(key, null));
}

static long resolveTerminationGracePeriodSeconds(Map<String, String> properties) {
long shutdownTimeoutMs =
PropertyUtil.propertyAsLong(
properties,
OptimizerProperties.OPTIMIZER_SHUTDOWN_TIMEOUT_MS,
OptimizerProperties.OPTIMIZER_SHUTDOWN_TIMEOUT_MS_DEFAULT);
long shutdownTimeoutSeconds = (shutdownTimeoutMs + 999L) / 1000L;
return shutdownTimeoutSeconds + TERMINATION_GRACE_BUFFER_SECONDS;
}

private KubernetesClient client;

@Override
Expand Down Expand Up @@ -118,6 +137,7 @@ protected Map<String, String> doScaleOut(Resource resource) {
String groupName = argsList.get("groupName").toString();
String resourceId = argsList.get("resourceId").toString();
String startUpArgs = argsList.get("startUpArgs").toString();
long terminationGraceSeconds = (long) argsList.get("terminationGraceSeconds");

String kubernetesName = NAME_PREFIX + resourceId;
Deployment deployment;
Expand All @@ -136,7 +156,8 @@ protected Map<String, String> doScaleOut(Resource resource) {
resourceId,
startUpArgs,
memory,
imagePullSecretsList);
imagePullSecretsList,
terminationGraceSeconds);
} else {
deployment =
initPodTemplateWithoutConfig(
Expand All @@ -147,7 +168,8 @@ protected Map<String, String> doScaleOut(Resource resource) {
resourceId,
startUpArgs,
memory,
imagePullSecretsList);
imagePullSecretsList,
terminationGraceSeconds);
}

client.apps().deployments().inNamespace(namespace).resource(deployment).create();
Expand Down Expand Up @@ -206,6 +228,11 @@ public Map<String, Object> generatePodStartArgs(
argsList.put("resourceId", resourceId);
argsList.put("groupName", groupName);
argsList.put("startUpArgs", startUpArgs);
// Derive the grace period from the same resource properties that drive the -st startup arg
// (buildOptimizerStartupArgsString), so the pod's SIGKILL deadline can never fall below the
// shutdown timeout the optimizer JVM actually honors.
argsList.put(
"terminationGraceSeconds", resolveTerminationGracePeriodSeconds(resource.getProperties()));

return argsList;
}
Expand All @@ -218,7 +245,8 @@ public Deployment initPodTemplateWithoutConfig(
String resourceId,
String startUpArgs,
long memory,
List<LocalObjectReference> imagePullSecretsList) {
List<LocalObjectReference> imagePullSecretsList,
long terminationGraceSeconds) {

DeploymentBuilder deploymentBuilder =
new DeploymentBuilder()
Expand All @@ -234,11 +262,12 @@ public Deployment initPodTemplateWithoutConfig(
.addToLabels("AmoroResourceId", resourceId)
.endMetadata()
.withNewSpec()
.withTerminationGracePeriodSeconds(terminationGraceSeconds)
.addNewContainer()
.withName("optimizer")
.withImage(image)
.withImagePullPolicy(pullPolicy)
.withCommand("sh", "-c", startUpArgs)
.withCommand("sh", "-c", "exec " + startUpArgs)
.withResources(
new ResourceRequirementsBuilder()
.withLimits(
Expand Down Expand Up @@ -288,7 +317,8 @@ public Deployment initPodTemplateFromFrontEnd(
String resourceId,
String startUpArgs,
long memory,
List<LocalObjectReference> imagePullSecretsList) {
List<LocalObjectReference> imagePullSecretsList,
long terminationGraceSeconds) {
podTemplate
.getTemplate()
.getMetadata()
Expand All @@ -305,7 +335,7 @@ public Deployment initPodTemplateFromFrontEnd(
container.setName("optimizer");
container.setImage(image);
container.setImagePullPolicy(pullPolicy);
container.setCommand(new ArrayList<>(Arrays.asList("sh", "-c", startUpArgs)));
container.setCommand(new ArrayList<>(Arrays.asList("sh", "-c", "exec " + startUpArgs)));

ResourceRequirements resourceRequirements = new ResourceRequirements();
resourceRequirements.setLimits(
Expand All @@ -324,6 +354,10 @@ public Deployment initPodTemplateFromFrontEnd(
podTemplate.getTemplate().getSpec().setImagePullSecrets(imagePullSecretsList);
}

if (podTemplate.getTemplate().getSpec().getTerminationGracePeriodSeconds() == null) {
podTemplate.getTemplate().getSpec().setTerminationGracePeriodSeconds(terminationGraceSeconds);
}

DeploymentSpec deploymentSpec = new DeploymentSpec();
deploymentSpec.setTemplate(podTemplate.getTemplate());

Expand Down
Loading
Loading