Package com.linkedin.venice.fastclient

Class RetriableAvroGenericStoreClient<K,V>

java.lang.Object

com.linkedin.venice.fastclient.InternalAvroStoreClient<K,V>

com.linkedin.venice.fastclient.DelegatingAvroStoreClient<K,V>

com.linkedin.venice.fastclient.RetriableAvroGenericStoreClient<K,V>

All Implemented Interfaces:

AvroGenericReadComputeStoreClient<K,V>, AvroGenericStoreClient<K,V>, Closeable, AutoCloseable

Direct Known Subclasses:

RetriableAvroSpecificStoreClient

public class RetriableAvroGenericStoreClient<K,V>
extends DelegatingAvroStoreClient<K,V>

This class is mostly used to trigger retry in the following scenarios: 1. The original request latency exceeds the retry threshold. 2. The original request fails. TODO: 1. Limit the retry volume. 2. Leverage some smart logic to avoid useless retry, such as retry triggered by heavy GC.

Field Summary

Fields

Modifier and Type	Field	Description
static final String	MULTI_KEY_LONG_TAIL_RETRY_STATS_PREFIX
static final String	SINGLE_KEY_LONG_TAIL_RETRY_STATS_PREFIX

Constructor Summary

Constructors

Constructor	Description
RetriableAvroGenericStoreClient(InternalAvroStoreClient<K,V> delegate, ClientConfig clientConfig, TimeoutProcessor timeoutProcessor)

Method Summary

Modifier and Type	Method	Description
void	close()	Release the internal resources.
protected void	compute(ComputeRequestContext<K,V> requestContext, ComputeRequestWrapper computeRequestWrapper, Set<K> keys, org.apache.avro.Schema resultSchema, StreamingCallback<K,ComputeGenericRecord> callback, long preRequestTimeInNS)
protected CompletableFuture<V>	get(GetRequestContext<K> requestContext, K key)	TODO: Limit the retry volume: Even though retry for a single request is being scheduled at max twice (once via scheduler (LONG_TAIL_RETRY) and once instant (ERROR_RETRY) if originalRequestFuture fails), there is no way to control the total allowed retry per node.
protected void	streamingBatchGet(BatchGetRequestContext<K,V> requestContext, Set<K> keys, StreamingCallback<K,V> callback)

Methods inherited from class com.linkedin.venice.fastclient.DelegatingAvroStoreClient

compute, getClientConfig, getKeySchema, getLatestValueSchema, getSchemaReader, getStoreName, start

Methods inherited from class com.linkedin.venice.fastclient.InternalAvroStoreClient

batchGet, batchGet, compute, computeWithKeyPrefixFilter, get, isProjectionFieldValidationEnabled, streamingBatchGet, streamingBatchGet, streamingBatchGet

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface com.linkedin.venice.client.store.AvroGenericReadComputeStoreClient

compute, compute, compute, computeAggregation

Methods inherited from interface com.linkedin.venice.client.store.AvroGenericStoreClient

get

Field Details

SINGLE_KEY_LONG_TAIL_RETRY_STATS_PREFIX

public static final String SINGLE_KEY_LONG_TAIL_RETRY_STATS_PREFIX

See Also:

• Constant Field Values

MULTI_KEY_LONG_TAIL_RETRY_STATS_PREFIX

public static final String MULTI_KEY_LONG_TAIL_RETRY_STATS_PREFIX

See Also:

• Constant Field Values

Constructor Details

RetriableAvroGenericStoreClient

public RetriableAvroGenericStoreClient(InternalAvroStoreClient<K,V> delegate,
 ClientConfig clientConfig,
 TimeoutProcessor timeoutProcessor)

Method Details

get

protected CompletableFuture<V> get(GetRequestContext<K> requestContext,
 K key)
                            throws VeniceClientException

TODO: Limit the retry volume: Even though retry for a single request is being scheduled at max twice (once via scheduler (LONG_TAIL_RETRY) and once instant (ERROR_RETRY) if originalRequestFuture fails), there is no way to control the total allowed retry per node. It would be good to design some mechanism to make it configurable, such as retry at most the slowest 5% of traffic, otherwise, too many retry requests could cause cascading failure.

Overrides:

get in class DelegatingAvroStoreClient<K,V>

Throws:

VeniceClientException

streamingBatchGet

protected void streamingBatchGet(BatchGetRequestContext<K,V> requestContext,
 Set<K> keys,
 StreamingCallback<K,V> callback)
                          throws VeniceClientException

Overrides:

streamingBatchGet in class DelegatingAvroStoreClient<K,V>

Throws:

VeniceClientException

compute

protected void compute(ComputeRequestContext<K,V> requestContext,
 ComputeRequestWrapper computeRequestWrapper,
 Set<K> keys,
 org.apache.avro.Schema resultSchema,
 StreamingCallback<K,ComputeGenericRecord> callback,
 long preRequestTimeInNS)
                throws VeniceClientException

Overrides:

compute in class DelegatingAvroStoreClient<K,V>

Throws:

VeniceClientException

close

public void close()

Description copied from interface: AvroGenericStoreClient

Release the internal resources.

Specified by:

close in interface AutoCloseable

Specified by:

close in interface AvroGenericStoreClient<K,V>

Specified by:

close in interface Closeable

Overrides:

close in class DelegatingAvroStoreClient<K,V>