Handle concurrent write conflicts in Apache Iceberg on the AWS Glue Knowledge Catalog

In trendy knowledge architectures, Apache Iceberg has emerged as a preferred desk format for knowledge lakes, providing key options together with ACID transactions and concurrent write assist. Though these capabilities are highly effective, implementing them successfully in manufacturing environments presents distinctive challenges that require cautious consideration.

Take into account a standard situation: A streaming pipeline repeatedly writes knowledge to an Iceberg desk whereas scheduled upkeep jobs carry out compaction operations. Though Iceberg gives built-in mechanisms to deal with concurrent writes, sure battle eventualities—corresponding to between streaming updates and compaction operations—can result in transaction failures that require particular dealing with patterns.

This submit demonstrates easy methods to implement dependable concurrent write dealing with mechanisms in Iceberg tables. We are going to discover Iceberg’s concurrency mannequin, study frequent battle eventualities, and supply sensible implementation patterns of each automated retry mechanisms and conditions requiring customized battle decision logic for constructing resilient knowledge pipelines. We may even cowl the sample with automated compaction by means of AWS Glue Knowledge Catalog desk optimization.

Frequent battle eventualities

Probably the most frequent knowledge conflicts happen in a number of particular operational eventualities that many organizations encounter of their knowledge pipelines, which we talk about on this part.

Concurrent UPDATE/DELETE on overlapping partitions

When a number of processes try to switch the identical partition concurrently, knowledge conflicts can come up. For instance, think about a knowledge high quality course of updating buyer information with corrected addresses whereas one other course of is deleting outdated buyer information. Each operations goal the identical partition primarily based on customer_id, resulting in potential conflicts as a result of they’re modifying an overlapping dataset. These conflicts are notably frequent in large-scale knowledge cleanup operations.

Compaction vs. streaming writes

A traditional battle situation happens throughout desk upkeep operations. Take into account a streaming pipeline ingesting real-time occasion knowledge whereas a scheduled compaction job runs to optimize file sizes. The streaming course of may be writing new information to a partition whereas the compaction job is making an attempt to mix present recordsdata in the identical partition. This situation is very frequent with Knowledge Catalog desk optimization, the place automated compaction can run concurrently with steady knowledge ingestion.

Concurrent MERGE operations

MERGE operations are notably inclined to conflicts as a result of they contain each studying and writing knowledge. As an illustration, an hourly job may be merging buyer profile updates from a supply system whereas a separate job is merging choice updates from one other system. If each jobs try to switch the identical buyer information, they will battle as a result of every operation bases its modifications on a unique view of the present knowledge state.

Basic concurrent desk updates

When a number of transactions happen concurrently, some transactions would possibly fail to decide to the catalog on account of interference from different transactions. Iceberg has mechanisms to deal with this situation, so it could adapt to concurrent transactions in lots of circumstances. Nonetheless, commits can nonetheless fail if the newest metadata is up to date after the bottom metadata model is established. This situation applies to any kind of updates on an Iceberg desk.

Iceberg’s concurrency mannequin and battle kind

Earlier than diving into particular implementation patterns, it’s important to know how Iceberg manages concurrent writes by means of its desk structure and transaction mannequin. Iceberg makes use of a layered structure to handle desk state and knowledge:

• Catalog layer – Maintains a pointer to the present desk metadata file, serving as the one supply of reality for desk state. The Knowledge Catalog gives the performance because the Iceberg catalog.
• Metadata layer – Accommodates metadata recordsdata that observe desk historical past, schema evolution, and snapshot data. These recordsdata are saved on Amazon Easy Storage Service (Amazon S3).
• Knowledge layer – Shops the precise knowledge recordsdata and delete recordsdata (for Merge-on-Learn operations). These recordsdata are additionally saved on Amazon S3.

The next diagram illustrates this structure.

This structure is prime to Iceberg’s optimistic concurrency management, the place a number of writers can proceed with their operations concurrently, and conflicts are detected at commit time.

Write transaction circulate

A typical write transaction in Iceberg follows these key steps:

1. Learn present state. In lots of operations (like OVERWRITE, MERGE, and DELETE), the question engine must know which recordsdata or rows are related, so it reads the present desk snapshot. That is non-obligatory for operations like INSERT.
2. Decide the modifications in transaction, and write new knowledge recordsdata.
3. Load the desk’s newest metadata, and decide which metadata model is used as the bottom for the replace.
4. Verify if the change ready in Step 2 is appropriate with the newest desk knowledge in Step 3. If the verify failed, the transaction should cease.
5. Generate new metadata recordsdata.
6. Commit the metadata recordsdata to the catalog. If the commit failed, retry from Step 3. The variety of retries relies on the configuration.

The next diagram illustrates this workflow.

Conflicts can happen at two important factors:

• Knowledge replace conflicts – Throughout validation when checking for knowledge conflicts (Step 4)
• Catalog commit conflicts – In the course of the commit when making an attempt to replace the catalog pointer (Step 6)

When working with Iceberg tables, understanding the sorts of conflicts that may happen and the way they’re dealt with is essential for constructing dependable knowledge pipelines. Let’s study the 2 major sorts of conflicts and their traits.

Catalog commit conflicts

Catalog commit conflicts happen when a number of writers try to replace the desk metadata concurrently. When a commit battle happens, Iceberg will routinely retry the operation primarily based on the desk’s write properties. The retry course of solely repeats the metadata commit, not your complete transaction, making it each secure and environment friendly. When the retries fail, the transaction fails with CommitFailedException.

Within the following diagram, two transactions run concurrently. Transaction 1 efficiently updates the desk’s newest snapshot within the Iceberg catalog from 0 to 1. In the meantime, transaction 2 makes an attempt to replace from Snapshot 0 to 1, however when it tries to commit the modifications to the catalog, it finds that the newest snapshot has already been modified to 1 by transaction 1. Consequently, transaction 2 must retry from Step 3.

These conflicts are sometimes transient and could be routinely resolved by means of retries. You may optionally configure write properties controlling commit retry habits. For extra detailed configuration, check with Write properties within the Iceberg documentation.

The metadata used when studying the present state (Step 1) and the snapshot used as base metadata for updates (Step 3) could be totally different. Even when one other transaction updates the newest snapshot between Steps 1 and three, the present transaction can nonetheless commit modifications to the catalog so long as it passes the info battle verify (Step 4). Which means that even when computing modifications and writing knowledge recordsdata (Step 1 to 2) take a very long time, and different transactions make modifications throughout this era, the transaction can nonetheless try to decide to the catalog. This demonstrates Iceberg’s clever concurrency management mechanism.

The next diagram illustrates this workflow.

Knowledge replace conflicts

Knowledge replace conflicts are extra advanced and happen when concurrent transactions try to switch overlapping knowledge. Throughout a write transaction, the question engine checks consistency between the snapshot being written and the newest snapshot in keeping with transaction isolation guidelines. When incompatibility is detected, the transaction fails with a ValidationException.

Within the following diagram, two transactions run concurrently on an worker desk containing id, identify, and wage columns. Transaction 1 makes an attempt to replace a document primarily based on Snapshot 0 and efficiently commits this alteration, making the newest snapshot model 1. In the meantime, transaction 2 additionally makes an attempt to replace the identical document primarily based on Snapshot 0. When transaction 2 initially scanned the info, the newest snapshot was 0, but it surely has since been up to date to 1 by transaction 1. In the course of the knowledge battle verify, transaction 2 discovers that its modifications battle with Snapshot 1, ensuing within the transaction failing.

These conflicts can’t be routinely retried by Iceberg’s library as a result of when knowledge conflicts happen, the desk’s state has modified, making it unsure whether or not retrying the transaction would keep general knowledge consistency. You could deal with this kind of battle primarily based in your particular use case and necessities.

The next desk summarizes how totally different write patterns have various probability of conflicts.

Iceberg desk’s isolation ranges

Iceberg tables assist two isolation ranges: Serializable and Snapshot isolation. Each present a learn constant view of the desk and guarantee readers see solely dedicated knowledge. Serializable isolation ensures that concurrent operations run as in the event that they had been carried out in some sequential order. Snapshot isolation gives weaker ensures however gives higher efficiency in environments with many concurrent writers. Underneath snapshot isolation, knowledge battle checks can cross even when concurrent transactions add new recordsdata with information that probably match its situations.

By default, Iceberg tables use serializable isolation. You may configure isolation ranges for particular operations utilizing desk properties:

tbl_properties = {
    'write.delete.isolation-level' = 'serializable',
    'write.replace.isolation-level' = 'serializable',
    'write.merge.isolation-level' = 'serializable'
}

You have to select the suitable isolation degree primarily based in your use case. Be aware that for conflicts between streaming ingestion and compaction operations, which is likely one of the most typical eventualities, snapshot isolation doesn’t present any further advantages to the default serializable isolation. For extra detailed configuration, see IsolationLevel.

Implementation patterns

Implementing strong concurrent write dealing with in Iceberg requires totally different methods relying on the battle kind and use case. On this part, we share confirmed patterns for dealing with frequent eventualities.

Handle catalog commit conflicts

Catalog commit conflicts are comparatively easy to deal with by means of desk properties. The next configurations function preliminary baseline settings you can modify primarily based in your particular workload patterns and necessities.

For frequent concurrent writes (for instance, streaming ingestion):

tbl_properties = {
    'commit.retry.num-retries': '10',
    'commit.retry.min-wait-ms': '100',
    'commit.retry.max-wait-ms': '10000',
    'commit.retry.total-timeout-ms': '1800000'
}

For upkeep operations (for instance, compaction):

tbl_properties = {
    'commit.retry.num-retries': '4',
    'commit.retry.min-wait-ms': '1000',
    'commit.retry.max-wait-ms': '60000',
    'commit.retry.total-timeout-ms': '1800000'
}

Handle knowledge replace conflicts

For knowledge replace conflicts, which may’t be routinely retried, you could implement a customized retry mechanism with correct error dealing with. A typical situation is when stream UPSERT ingestion conflicts with concurrent compaction operations. In such circumstances, the stream ingestion job ought to sometimes implement retries to deal with incoming knowledge. With out correct error dealing with, the job will fail with a ValidationException.

We present two instance scripts demonstrating a sensible implementation of error dealing with for knowledge conflicts in Iceberg streaming jobs. The code particularly catches ValidationException by means of Py4JJavaError dealing with, which is crucial for correct Java-Python interplay. It consists of exponential backoff and jitter technique by including a random delay of 0–25% to every retry interval. For instance, if the bottom exponential backoff time is 4 seconds, the precise retry delay might be between 4–5 seconds, serving to stop rapid retry storms whereas sustaining affordable latency.

On this instance, we create a situation with frequent MERGE operations on the identical information through the use of 'worth' as a novel identifier and artificially limiting its vary. By making use of a modulo operation (worth % 20), we constrain all values to fall inside 0–19, which suggests a number of updates will goal the identical information. As an illustration, if the unique stream comprises values 0, 20, 40, and 60, they are going to all be mapped to 0, leading to a number of MERGE operations focusing on the identical document. We then use groupBy and max aggregation to simulate a typical UPSERT sample the place we maintain the newest document for every worth. The reworked knowledge is saved in a short lived view that serves because the supply desk within the MERGE assertion, permitting us to carry out UPDATE operations utilizing 'worth' because the matching situation. This setup helps exhibit how our retry mechanism handles ValidationExceptions that happen when concurrent transactions try to switch the identical information.

The primary instance makes use of Spark Structured Streaming utilizing a charge supply with a 20-second set off interval to exhibit the retry mechanism’s habits when concurrent operations trigger knowledge conflicts. Substitute together with your database identify, together with your desk identify, amzn-s3-demo-bucket together with your S3 bucket identify.

import time
import random
from pyspark.sql import SparkSession
from py4j.protocol import Py4JJavaError
from pyspark.sql.features import max as max_

CATALOG = "glue_catalog"
DATABASE = ""
TABLE = ""
BUCKET = "amzn-s3-demo-bucket"

spark = SparkSession.builder 
    .appName("IcebergUpsertExample") 
    .config(f"spark.sql.catalog.{CATALOG}", "org.apache.iceberg.spark.SparkCatalog") 
    .config("spark.sql.extensions","org.apache.iceberg.spark.extensions.IcebergSparkSessionExtensions") 
    .config(f"spark.sql.catalog.{CATALOG}.io-impl","org.apache.iceberg.aws.s3.S3FileIO") 
    .config("spark.sql.defaultCatalog", CATALOG) 
    .config(f"spark.sql.catalog.{CATALOG}.kind", "glue") 
    .getOrCreate()
    
spark.sql(f"""
    CREATE TABLE IF NOT EXISTS {DATABASE}.{TABLE} (
        timestamp TIMESTAMP,
        worth LONG
    )
    USING iceberg
    LOCATION 's3://{BUCKET}/warehouse'
""")

def backoff(try):
    """Exponential backoff with jitter"""
    exp_backoff = min(2 ** try, 60)
    jitter = random.uniform(0, 0.25 * exp_backoff)
    return exp_backoff + jitter

def is_validation_exception(java_exception):
    """Verify if exception is ValidationException"""
    trigger = java_exception
    whereas trigger shouldn't be None:
        if "org.apache.iceberg.exceptions.ValidationException" in str(trigger.getClass().getName()):
            return True
        trigger = trigger.getCause()
    return False

def upsert_with_retry(microBatchDF, batchId):
    max_retries = 5
    try = 0
    
    # Use a narrower key vary to deliberately improve updates for a similar worth in MERGE
    transformedDF = microBatchDF 
        .selectExpr("timestamp", "worth % 20 AS worth") 
        .groupBy("worth") 
        .agg(max_("timestamp").alias("timestamp"))
        
    view_name = f"incoming_data_{batchId}"
    transformedDF.createOrReplaceGlobalTempView(view_name)
    
    whereas try 

The second instance makes use of GlueContext.forEachBatch accessible on AWS Glue Streaming jobs. The implementation sample for the retry mechanism stays the identical, however the primary variations are the preliminary setup utilizing GlueContext and easy methods to create a streaming DataFrame. Though our instance makes use of spark.readStream with a charge supply for demonstration, in precise AWS Glue Streaming jobs, you’ll sometimes create your streaming DataFrame utilizing glueContext.create_data_frame.from_catalog to learn from sources like Amazon Kinesis or Kafka. For extra particulars, see AWS Glue Streaming connections. Substitute together with your database identify, together with your desk identify, amzn-s3-demo-bucket together with your S3 bucket identify.

import time
import random
from py4j.protocol import Py4JJavaError
from pyspark.context import SparkContext
from awsglue.context import GlueContext
from pyspark.sql import SparkSession
from pyspark.sql.features import max as max_

CATALOG = "glue_catalog"
DATABASE = ""
TABLE = ""
BUCKET = "amzn-s3-demo-bucket"

spark = SparkSession.builder 
    .appName("IcebergUpsertExample") 
    .config(f"spark.sql.catalog.{CATALOG}", "org.apache.iceberg.spark.SparkCatalog") 
    .config("spark.sql.extensions","org.apache.iceberg.spark.extensions.IcebergSparkSessionExtensions") 
    .config(f"spark.sql.catalog.{CATALOG}.io-impl","org.apache.iceberg.aws.s3.S3FileIO") 
    .config("spark.sql.defaultCatalog", CATALOG) 
    .config(f"spark.sql.catalog.{CATALOG}.kind", "glue") 
    .getOrCreate()

sc = spark.sparkContext
glueContext = GlueContext(sc)

spark.sql(f"""
    CREATE TABLE IF NOT EXISTS {DATABASE}.{TABLE} (
        timestamp TIMESTAMP,
        worth LONG
    )
    USING iceberg
    LOCATION 's3://{BUCKET}/warehouse'
""")

def backoff(try):
    exp_backoff = min(2 ** try, 60)
    jitter = random.uniform(0, 0.25 * exp_backoff)
    return exp_backoff + jitter

def is_validation_exception(java_exception):
    trigger = java_exception
    whereas trigger shouldn't be None:
        if "org.apache.iceberg.exceptions.ValidationException" in str(trigger.getClass().getName()):
            return True
        trigger = trigger.getCause()
    return False

def upsert_with_retry(batch_df, batchId):
    max_retries = 5
    try = 0
    transformedDF = batch_df.selectExpr("timestamp", "worth % 20 AS worth") 
                           .groupBy("worth") 
                           .agg(max_("timestamp").alias("timestamp"))
                           
    view_name = f"incoming_data_{batchId}"
    transformedDF.createOrReplaceGlobalTempView(view_name)
    
    whereas try 

Reduce battle risk by scoping your operations

When performing upkeep operations like compaction or updates, it’s really useful to slender down the scope to attenuate overlap with different operations. For instance, take into account a desk partitioned by date the place a streaming job repeatedly upserts knowledge for the newest date. The next is the instance script to run the rewrite_data_files process to compact your complete desk:

# Instance of broad scope compaction
spark.sql("""
   CALL catalog_name.system.rewrite_data_files(
       desk => 'db.table_name'
   )
""")

By narrowing the compaction scope with a date partition filter within the the place clause, you possibly can keep away from conflicts between streaming ingestion and compaction operations. The streaming job can proceed to work with the newest partition whereas compaction processes historic knowledge.

# Slim down the scope by partition
spark.sql("""
    CALL catalog_name.system.rewrite_data_files(
        desk => 'db.table_name',
        the place => 'date_partition 

Conclusion

Efficiently managing concurrent writes in Iceberg requires understanding each the desk structure and varied battle eventualities. On this submit, we explored easy methods to implement dependable battle dealing with mechanisms in manufacturing environments.

Probably the most important idea to recollect is the excellence between catalog commit conflicts and knowledge conflicts. Though catalog commit conflicts could be dealt with by means of automated retries and desk properties configuration, knowledge conflicts require cautious implementation of customized dealing with logic. This turns into notably vital when implementing upkeep operations like compaction, the place utilizing the the place clause in rewrite_data_files can considerably reduce battle potential by lowering the scope of operations.

For streaming pipelines, the important thing to success lies in implementing correct error dealing with that may differentiate between battle varieties and reply appropriately. This consists of configuring appropriate retry settings by means of desk properties and implementing backoff methods that align together with your workload traits. When mixed with well-timed upkeep operations, these patterns assist construct resilient knowledge pipelines that may deal with concurrent writes reliably.

By making use of these patterns and understanding the underlying mechanisms of Iceberg’s concurrency mannequin, you possibly can construct strong knowledge pipelines that successfully deal with concurrent write eventualities whereas sustaining knowledge consistency and reliability.

In regards to the Authors

Sotaro Hikita is an Analytics Options Architect. He helps clients throughout a variety of industries in constructing and working analytics platforms extra successfully. He’s notably enthusiastic about massive knowledge applied sciences and open supply software program.

Noritaka Sekiyama is a Principal Large Knowledge Architect on the AWS Glue group. He works primarily based in Tokyo, Japan. He’s accountable for constructing software program artifacts to assist clients. In his spare time, he enjoys biking together with his street bike.