Skip to content

[ML] address potential bug where trained models get stuck in starting after being allocated to node #88945

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
merged 5 commits into from
Aug 1, 2022
Merged

[ML] address potential bug where trained models get stuck in starting after being allocated to node #88945

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
67dc243
[ML] address potential bug where trained models get stuck in starting
benwtrent Jul 29, 2022
9c30dfa
Update docs/changelog/88945.yaml
benwtrent Jul 29, 2022
b254d5e
Merge branch 'main' into bugfix/ml-address-potential-model-start-thre…
elasticmachine Jul 29, 2022
be745a4
Update x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/infe…
benwtrent Aug 1, 2022
ab5f0b4
Merge branch 'main' into bugfix/ml-address-potential-model-start-thre…
elasticmachine Aug 1, 2022
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
6 changes: 6 additions & 0 deletions docs/changelog/88945.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,6 @@
pr: 88945
summary: Address potential bug where trained models get stuck in starting after being
allocated to node
area: Machine Learning
type: bug
issues: []
7 changes: 3 additions & 4 deletions ...va/org/elasticsearch/xpack/ml/inference/assignment/TrainedModelAssignmentNodeService.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -225,14 +225,13 @@ void loadQueuedModels() {
} catch (Exception ex) {
logger.warn(() -> "[" + modelId + "] Start deployment failed", ex);
if (ExceptionsHelper.unwrapCause(ex) instanceof ResourceNotFoundException) {
logger.warn(() -> "[" + modelId + "] Start deployment failed", ex);
logger.debug(() -> "[" + modelId + "] Start deployment failed as model was not found", ex);
handleLoadFailure(loadingTask, ExceptionsHelper.missingTrainedModel(modelId, ex));
} else if (ExceptionsHelper.unwrapCause(ex) instanceof SearchPhaseExecutionException) {
logger.trace(() -> "[" + modelId + "] Start deployment failed, will retry", ex);
logger.debug(() -> "[" + modelId + "] Start deployment failed, will retry", ex);
// A search phase execution failure should be retried, push task back to the queue
loadingToRetry.add(loadingTask);
} else {
logger.warn(() -> "[" + modelId + "] Start deployment failed", ex);
handleLoadFailure(loadingTask, ex);
}
}
Expand Down Expand Up @@ -413,7 +412,7 @@ private void updateNumberOfAllocations(TrainedModelAssignmentMetadata assignment
for (TrainedModelAssignment assignment : modelsToUpdate) {
TrainedModelDeploymentTask task = modelIdToTask.get(assignment.getModelId());
if (task == null) {
logger.debug(() -> format("[%s] task was removed whilst updating number of allocations", task.getModelId()));
logger.debug(() -> format("[%s] task was removed whilst updating number of allocations", assignment.getModelId()));
continue;
}
RoutingInfo routingInfo = assignment.getNodeRoutingTable().get(nodeId);
Expand Down
73 changes: 44 additions & 29 deletions ...n/ml/src/main/java/org/elasticsearch/xpack/ml/inference/deployment/DeploymentManager.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,6 +61,7 @@
import java.util.concurrent.atomic.AtomicLong;
import java.util.function.Consumer;

import static org.elasticsearch.core.Strings.format;
import static org.elasticsearch.xpack.core.ClientHelper.ML_ORIGIN;
import static org.elasticsearch.xpack.core.ClientHelper.executeAsyncWithOrigin;

Expand Down Expand Up @@ -149,34 +150,46 @@ private void doStartDeployment(TrainedModelDeploymentTask task, ActionListener<T
TrainedModelConfig modelConfig = getModelResponse.getResources().results().get(0);
processContext.modelInput.set(modelConfig.getInput());

assert modelConfig.getInferenceConfig() instanceof NlpConfig;
NlpConfig nlpConfig = (NlpConfig) modelConfig.getInferenceConfig();
task.init(nlpConfig);

SearchRequest searchRequest = vocabSearchRequest(nlpConfig.getVocabularyConfig(), modelConfig.getModelId());
executeAsyncWithOrigin(client, ML_ORIGIN, SearchAction.INSTANCE, searchRequest, ActionListener.wrap(searchVocabResponse -> {
if (searchVocabResponse.getHits().getHits().length == 0) {
listener.onFailure(
new ResourceNotFoundException(
Messages.getMessage(
Messages.VOCABULARY_NOT_FOUND,
task.getModelId(),
VocabularyConfig.docId(modelConfig.getModelId())
if (modelConfig.getInferenceConfig()instanceof NlpConfig nlpConfig) {
task.init(nlpConfig);

SearchRequest searchRequest = vocabSearchRequest(nlpConfig.getVocabularyConfig(), modelConfig.getModelId());
executeAsyncWithOrigin(client, ML_ORIGIN, SearchAction.INSTANCE, searchRequest, ActionListener.wrap(searchVocabResponse -> {
if (searchVocabResponse.getHits().getHits().length == 0) {
listener.onFailure(
new ResourceNotFoundException(
Messages.getMessage(
Messages.VOCABULARY_NOT_FOUND,
task.getModelId(),
VocabularyConfig.docId(modelConfig.getModelId())
)
)
)
);
return;
}

Vocabulary vocabulary = parseVocabularyDocLeniently(searchVocabResponse.getHits().getAt(0));
NlpTask nlpTask = new NlpTask(nlpConfig, vocabulary);
NlpTask.Processor processor = nlpTask.createProcessor();
processContext.nlpTaskProcessor.set(processor);
// here, we are being called back on the searching thread, which MAY be a network thread
// `startAndLoad` creates named pipes, blocking the calling thread, better to execute that in our utility
// executor.
executorServiceForDeployment.execute(
() -> startAndLoad(processContext, modelConfig.getLocation(), modelLoadedListener)
);
return;
}

Vocabulary vocabulary = parseVocabularyDocLeniently(searchVocabResponse.getHits().getAt(0));
NlpTask nlpTask = new NlpTask(nlpConfig, vocabulary);
NlpTask.Processor processor = nlpTask.createProcessor();
processContext.nlpTaskProcessor.set(processor);
// here, we are being called back on the searching thread, which MAY be a network thread
// `startAndLoad` creates named pipes, blocking the calling thread, better to execute that in our utility
// executor.
executorServiceForDeployment.execute(() -> startAndLoad(processContext, modelConfig.getLocation(), modelLoadedListener));
}, listener::onFailure));
}, listener::onFailure));
} else {
listener.onFailure(
new IllegalArgumentException(
format(
"[%s] must be a pytorch model; found inference config of kind [%s]",
modelConfig.getModelId(),
modelConfig.getInferenceConfig().getWriteableName()
)
)
);
}
}, listener::onFailure);

executeAsyncWithOrigin(
Expand Down Expand Up @@ -404,10 +417,12 @@ private Consumer<String> onProcessCrash() {
}

void loadModel(TrainedModelLocation modelLocation, ActionListener<Boolean> listener) {
if (modelLocation instanceof IndexLocation) {
process.get().loadModel(task.getModelId(), ((IndexLocation) modelLocation).getIndexName(), stateStreamer, listener);
if (modelLocation instanceof IndexLocation indexLocation) {
process.get().loadModel(task.getModelId(), indexLocation.getIndexName(), stateStreamer, listener);
} else {
throw new IllegalStateException("unsupported trained model location [" + modelLocation.getClass().getSimpleName() + "]");
listener.onFailure(
new IllegalStateException("unsupported trained model location [" + modelLocation.getClass().getSimpleName() + "]")
);
}
}

Expand Down
27 changes: 18 additions & 9 deletions ...in/java/org/elasticsearch/xpack/ml/inference/persistence/ChunkedTrainedModelRestorer.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,11 +10,11 @@
import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.Logger;
import org.elasticsearch.ResourceNotFoundException;
import org.elasticsearch.action.ActionListener;
import org.elasticsearch.action.search.SearchAction;
import org.elasticsearch.action.search.SearchRequest;
import org.elasticsearch.action.search.SearchRequestBuilder;
import org.elasticsearch.action.search.SearchResponse;
import org.elasticsearch.client.internal.Client;
import org.elasticsearch.client.internal.OriginSettingClient;
import org.elasticsearch.common.bytes.BytesReference;
import org.elasticsearch.common.xcontent.LoggingDeprecationHandler;
import org.elasticsearch.core.CheckedFunction;
Expand All @@ -38,8 +38,10 @@
import java.util.concurrent.ExecutorService;
import java.util.function.Consumer;

import static org.elasticsearch.core.Strings.format;
import static org.elasticsearch.xpack.core.ClientHelper.ML_ORIGIN;
import static org.elasticsearch.xpack.core.ClientHelper.executeAsyncWithOrigin;
import static org.elasticsearch.xpack.ml.MachineLearning.NATIVE_INFERENCE_COMMS_THREAD_POOL_NAME;
import static org.elasticsearch.xpack.ml.MachineLearning.UTILITY_THREAD_POOL_NAME;

/**
* Searches for and emits {@link TrainedModelDefinitionDoc}s in
Expand Down Expand Up @@ -71,7 +73,7 @@ public ChunkedTrainedModelRestorer(
ExecutorService executorService,
NamedXContentRegistry xContentRegistry
) {
this.client = client;
this.client = new OriginSettingClient(client, ML_ORIGIN);
this.executorService = executorService;
this.xContentRegistry = xContentRegistry;
this.modelId = modelId;
Expand Down Expand Up @@ -122,7 +124,6 @@ public void restoreModelDefinition(

logger.debug("[{}] restoring model", modelId);
SearchRequest searchRequest = buildSearch(client, modelId, index, searchSize, null);

executorService.execute(() -> doSearch(searchRequest, modelConsumer, successConsumer, errorConsumer));
}

Expand All @@ -132,8 +133,16 @@ private void doSearch(
Consumer<Boolean> successConsumer,
Consumer<Exception> errorConsumer
) {

executeAsyncWithOrigin(client, ML_ORIGIN, SearchAction.INSTANCE, searchRequest, ActionListener.wrap(searchResponse -> {
try {
assert Thread.currentThread().getName().contains(NATIVE_INFERENCE_COMMS_THREAD_POOL_NAME)
|| Thread.currentThread().getName().contains(UTILITY_THREAD_POOL_NAME)
: format(
"Must execute from [%s] or [%s] but thread is [%s]",
NATIVE_INFERENCE_COMMS_THREAD_POOL_NAME,
UTILITY_THREAD_POOL_NAME,
Thread.currentThread().getName()
);
SearchResponse searchResponse = client.search(searchRequest).actionGet();
if (searchResponse.getHits().getHits().length == 0) {
errorConsumer.accept(new ResourceNotFoundException(Messages.getMessage(Messages.MODEL_DEFINITION_NOT_FOUND, modelId)));
return;
Expand Down Expand Up @@ -182,13 +191,13 @@ private void doSearch(
searchRequestBuilder.searchAfter(new Object[] { lastHit.getIndex(), lastNum });
executorService.execute(() -> doSearch(searchRequestBuilder.request(), modelConsumer, successConsumer, errorConsumer));
}
}, e -> {
} catch (Exception e) {
if (ExceptionsHelper.unwrapCause(e) instanceof ResourceNotFoundException) {
errorConsumer.accept(new ResourceNotFoundException(Messages.getMessage(Messages.MODEL_DEFINITION_NOT_FOUND, modelId)));
} else {
errorConsumer.accept(e);
}
}));
}
}

private static SearchRequestBuilder buildSearchBuilder(Client client, String modelId, String index, int searchSize) {
Expand Down