Closed marc0-p closed 10 months ago
bryanck
commented
10 months ago Yes, that's the object store file layout feature of Iceberg. See https://iceberg.apache.org/docs/latest/aws/#object-store-file-layout for more info. You can disable it but I'd leave it on for the reasons given in the docs.
marc0-p
commented
10 months ago @bryanck - I think I found something.
So, in the object store file layout documentation they show a path like this:
s3://my-table-data-bucket/2d3905f8/my_ns.db/my_table/category=orders/00000-0-5affc076-96a4-48f2-9cd2-d5efbc9f0c94-00001.parquet
Where as my data path looks something like this
s3://my-bucket/datalake/topics/my_topic/data/2d3905f8/category=orders/00000-0-5affc076-96a4-48f2-9cd2-d5efbc9f0c94-00001.parquet
where my config would have the equivalent of:
iceberg.catalog.warehouse = s3a://my-bucket/datalake
topics.dir = topics
I think with my deterministic hash being inserted after the table name, might be a problem? Also, the apache example is a bit weird too, as the data is not under /data (counterpart to /metadata) in standard iceberg data structure.
marc0-p
commented
10 months ago Another thing. I'm not getting any manifest lists, or updates to the metadata. Which is probably the true root cause of all this!
So the status is, I have data actively flowing into the data dir, but have no updates to the metadata.
marc0-p
commented
10 months ago Note, while the connector is "running" I see this task error.:
{
"name": "sink-s3-block-ingestion-block-all-iceberg-v02",
"connector": {
"state": "RUNNING",
"worker_id": "XXXXXX:8083"
},
"tasks": [
{
"id": 0,
"state": "FAILED",
"worker_id": "XXXXXX:8083",
"trace": "java.lang.RuntimeException: java.util.concurrent.ExecutionException: org.apache.kafka.common.errors.UnknownTopicOrPartitionException: This server does not host this topic-partition.\n\tat io.tabular.iceberg.connect.channel.Coordinator.lambda$getTotalPartitionCount$0(Coordinator.java:129)\n\tat java.base/java.util.stream.ReferencePipeline$4$1.accept(ReferencePipeline.java:212)\n\tat java.base/java.util.HashMap$ValueSpliterator.forEachRemaining(HashMap.java:1693)\n\tat java.base/java.util.stream.AbstractPipeline.copyInto(AbstractPipeline.java:484)\n\tat java.base/java.util.stream.AbstractPipeline.wrapAndCopyInto(AbstractPipeline.java:474)\n\tat java.base/java.util.stream.ReduceOps$ReduceOp.evaluateSequential(ReduceOps.java:913)\n\tat java.base/java.util.stream.AbstractPipeline.evaluate(AbstractPipeline.java:234)\n\tat java.base/java.util.stream.IntPipeline.reduce(IntPipeline.java:491)\n\tat java.base/java.util.stream.IntPipeline.sum(IntPipeline.java:449)\n\tat io.tabular.iceberg.connect.channel.Coordinator.getTotalPartitionCount(Coordinator.java:132)\n\tat io.tabular.iceberg.connect.channel.Coordinator.
bryanck
commented
10 months ago It looks like you may be running an older version of the sink, can you try with the latest version? (Note, some property names were changed if going from 0.4 -> 0.5). The task should fail if the coordinator throws an exception so I'll make sure that's the case.
marcopalmeri
commented
10 months ago Hi @bryanck upgrading the version now results in a different error:
"org.apache.kafka.connect.errors.ConnectException: No partitions assigned, cannot determine leader\n\tat io.tabular.iceberg.connect.IcebergSinkTask.lambda$isLeader$1(IcebergSinkTask.java:109)\n\tat java.base/java.util.Optional.orElseThrow(Optional.java:408)\n\tat io.tabular.iceberg.connect.IcebergSinkTask.isLeader(IcebergSinkTask.java:108)\n\tat io.tabular.iceberg.connect.IcebergSinkTask.open(IcebergSinkTask.java:86)\n\tat org.apache.kafka.connect.runtime.WorkerSinkTask.openPartitions(WorkerSinkTask.java:637)\n\tat org.apache.kafka.connect.runtime.WorkerSinkTask.access$1200(WorkerSinkTask.java:72)\n\tat org.apache.kafka.connect.runtime.WorkerSinkTask$HandleRebalance.onPartitionsAssigned(WorkerSinkTask.java:733)\n\tat org.apache.kafka.clients.consumer.internals.ConsumerCoordinator.invokePartitionsAssigned(ConsumerCoordinator.java:317)\n\tat org.apache.kafka.clients.consumer.internals.ConsumerCoordinator.onJoinComplete(ConsumerCoordinator.java:466)\n\tat org.apache.kafka.clients.consumer.internals.AbstractCoordinator.joinGroupIfNeeded(AbstractCoordinator.java:474)\n\tat org.apache.kafka.clients.consumer.internals.AbstractCoordinator.ensureActiveGroup(AbstractCoordinator.java:385)\n\tat org.apache.kafka.clients.consumer.internals.ConsumerCoordinator.poll(ConsumerCoordinator.java:552)\n\tat org.apache.kafka.clients.consumer.KafkaConsumer.updateAssignmentMetadataIfNeeded(KafkaConsumer.java:1276)\n\tat org.apache.kafka.clients.consumer.KafkaConsumer.poll(KafkaConsumer.java:1240)\n\tat org.apache.kafka.clients.consumer.KafkaConsumer.poll(KafkaConsumer.java:1220)\n\tat org.apache.kafka.connect.runtime.WorkerSinkTask.pollConsumer(WorkerSinkTask.java:472)\n\tat org.apache.kafka.connect.runtime.WorkerSinkTask.poll(WorkerSinkTask.java:328)\n\tat org.apache.kafka.connect.runtime.WorkerSinkTask.iteration(WorkerSinkTask.java:234)\n\tat org.apache.kafka.connect.runtime.WorkerSinkTask.execute(WorkerSinkTask.java:203)\n\tat org.apache.kafka.connect.runtime.WorkerTask.doRun(WorkerTask.java:189)\n\tat org.apache.kafka.connect.runtime.WorkerTask.run(WorkerTask.java:244)\n\tat java.base/java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:515)\n\tat java.base/java.util.concurrent.FutureTask.run(FutureTask.java:264)\n\tat java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128)\n\tat java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628)\n\tat java.base/java.lang.Thread.run(Thread.java:829)\n"Any ideas? I tried bumping the connector version already.
bryanck
commented
10 months ago It looks like, as with the previous error, that the source topic cannot be found. Can you post your full config and how you are running Kafka Connect?
marcopalmeri
commented
10 months ago Using terraform, the same repo/deployments for all of our different sinks (mostly just confluent s3 sinks, some jdbc).
resource "kafka-connect_connector" "sink-iceberg-block-ingestion" {
name = "sink-s3-block-ingestion-block-all-iceberg-v02"
config = merge({
"name" = "sink-s3-block-ingestion-block-all-iceberg-v02"
"connector.class" = "io.tabular.iceberg.connect.IcebergSinkConnector"
"timezone" = "UTC"
"locale" = "en-GB"
"topics" = "prod.block-ingestion.block_data.block.all"
"topics.dir" = "iceberg-ingested/kafka-uat/topics"
"iceberg.tables" = "uat_ingested_block.block_all"
"tasks.max" = "2"
"iceberg.catalog.catalog-impl" = "org.apache.iceberg.aws.glue.GlueCatalog"
"iceberg.catalog" = "uat_ingested_block"
"iceberg.catalog.warehouse" = "s3a://blah/iceberg-ingested"
"iceberg.catalog.io-impl" = "org.apache.iceberg.aws.s3.S3FileIO"
"iceberg.catalog.client.region" = local.aws_region
"iceberg.tables.evolve-schema-enabled" = "true"
"iceberg.tables.upsert-mode-enabled" = "true"
"iceberg.tables.auto-create-enabled" = "true"
"iceberg.tables.default-id-columns" = "id"
"iceberg.tables.default-partition-by" = "base_chain"
# "iceberg.table.uat_ingested_block.block_all.id-columns" = "id"
# "iceberg.table.uat_ingested_block.block_all.partition-by" = "base_chain"
"value.converter" = "io.confluent.connect.avro.AvroConverter"
"value.converter.schema.registry.url" = var.schema_registry_url
"value.converter.value.subject.name.strategy" = "io.confluent.kafka.serializers.subject.RecordNameStrategy"
"key.converter.schema.registry.url" = var.schema_registry_url
"schema.compatibility" = "NONE"
"rotate.schedule.interval.ms" = "60000"
"errors.retry.delay.max.ms" = "5000"
"errors.retry.timeout" = "60000"
"consumer.override.auto.offset.reset" = "earliest"
"connect.meta.data" = "false"
"enhanced.avro.schema.support" = "true"
"behavior.on.null.values" = "ignore"
"consumer.override.max.partition.fetch.bytes" = "120000000"
"consumer.override.max.request.size" = "120000000"
"consumer.override.max.poll.records" = "20"
"consumer.override.fetch.min.bytes" = "500000"
"consumer.override.fetch.max.bytes" = "120000000"
"consumer.override.session.timeout.ms" = "200000"
"consumer.override.request.interval.ms" = "180000"
"consumer.override.max.poll.interval.ms" = "180000"
"consumer.override.max.message.bytes" = "100000000"
"consumer.override.fetch.max.message.bytes" = "100000000"
"consumer.override.receive.buffer.bytes" = "100000000"
"consumer.override.request.timeout.ms" = "90000"
"transforms" = "tombstoneHandler"
"transforms.tombstoneHandler.type" = "io.confluent.connect.transforms.TombstoneHandler"
"transforms.tombstoneHandler.behavior" = "ignore"
}, local.funding_conn_override) #These are the connection/ssl details.
}It could be that the sink is not picking up the correct Kafka settings. It will attempt to read them from the worker properties file. You can try setting Kafka properties for the sink's internal Kafka client explicitly by setting iceberg.kafka.* properties. The prefix for these properties will be stripped off and then they will be passed to the Kafka client during initialization.
bryanck
commented
10 months ago Another user reported a similar error where the leader can't be determined, so I'm taking a further look, I'll report back.
bryanck
commented
10 months ago I have a new build 0.5.8 that hopefully addresses the leader election failure you're seeing.
marcopalmeri
commented
10 months ago @bryanck We have no errors now, the sink is writing data to S3, but still no metadata updates, no manifest list, no snapshots.
Is this because we manually created the table?
The auto-create did not work for us.
bryanck
commented
10 months ago Unfortunately I’m not able to reproduce your issue, I’m able to auto create and insert using a Glue catalog. If you can give me a way to reproduce your issue with a simplified example I’d be happy to investigate.
bryanck
commented
10 months ago In case it helps, here's the config I used:
topics: test-topic
header.converter: org.apache.kafka.connect.storage.StringConverter
key.converter: org.apache.kafka.connect.storage.StringConverter
value.converter: org.apache.kafka.connect.json.JsonConverter
value.converter.schemas.enable: false
iceberg.tables: kc_glue_test.foobar
iceberg.table.kc_glue_test.foobar.partition-by: id
iceberg.tables.auto-create-enabled: true
iceberg.control.commit.interval-ms: 30000
iceberg.catalog.catalog-impl: org.apache.iceberg.aws.glue.GlueCatalog
iceberg.catalog.io-impl: org.apache.iceberg.aws.s3.S3FileIO
iceberg.catalog.warehouse: s3://kc-glue-test/iceberg
iceberg.catalog.s3.staging-dir: /home/kafka/iceberg-stagingI'm going to close this, feel free to open a new ticket if you have further issues and a method to reproduce.
Hi, I set up this sink connector and it's working but I have an issue.
I'm getting an extra level in the data directory, but before the partition folders. Almost like some kind of sharding. It looks something like this:
s3://my_bucket/topics/some.topic/data/00u8bQ/partition_field=xyz/part_file_000.parquet
Where the "00u8bQ" folder is one of many 6-character folders under the data directory.