Add `org.apache.iceberg.SnapshotUpdate#metricsReporter(MetricsReporter reporter)`

[raunaqmorarka]

Feature Request / Improvement

The Scan API has org.apache.iceberg.Scan#metricsReporter available to register a custom metrics reporter for collecting metrics from the scan planning process.
We need a similar API on the SnapshotUpdate interface to allow collecting org.apache.iceberg.metrics.CommitReport from a custom metrics reporter. This will allow engines like Trino to collect and display metrics about the work done during a table change operation.

Query engine

None

Willingness to contribute

• I can contribute this improvement/feature independently
• I would be willing to contribute this improvement/feature with guidance from the Iceberg community
• I cannot contribute this improvement/feature at this time

[raunaqmorarka]

fyi @ebyhr @findepi @findinpath

[singhpk234]

[for my understanding] can't we just pass the reporter when the table obj is created so that it percolates down to SnapshotProducer which can capture the metrics ?

iceberg/core/src/main/java/org/apache/iceberg/SnapshotProducer.java

Lines 520 to 529 in f17f34b

	reporter.report(
	ImmutableCommitReport.builder()
	.tableName(createSnapshotEvent.tableName())
	.snapshotId(createSnapshotEvent.snapshotId())
	.operation(createSnapshotEvent.operation())
	.sequenceNumber(createSnapshotEvent.sequenceNumber())
	.metadata(EnvironmentContext.get())
	.commitMetrics(
	CommitMetricsResult.from(commitMetrics(), createSnapshotEvent.summary()))
	.build());

Nevertheless please feel to open PR, your contributions are very welcomed !

[raunaqmorarka]

@singhpk234 it is indeed possible to pass in metrics reporter as a constructor argument in org.apache.iceberg.BaseTable#BaseTable(org.apache.iceberg.TableOperations, java.lang.String, org.apache.iceberg.metrics.MetricsReporter). I think we can use this by creating new BaseTable objects with our metrics reporter where needed, but the API is not as nice and it looks like this would replace any existing metrics reporter in BaseTable.
E.g. In case of REST catalog, the MetricsReporter is already set by org.apache.iceberg.rest.RESTSessionCatalog and we would like to plug in our custom reporter in addition to that rather than replacing it.

[nastra]

@raunaqmorarka how exactly are you initializing the metrics reporter in your case? We use MetricsReporters.combine(..) which guarantees that metrics reporting is done via REST and via the reporter that was originally configured

[manuzhang]

I have to add a combineMetricsReporter to BaseTable to implement write metrics for Spark. It would be nice to have a more general API.

[raunaqmorarka]

I found that while we can add metrics reporter to org.apache.iceberg.BaseTable, the same is not possible for org.apache.iceberg.BaseTransaction#BaseTransaction because that does not expose the type and reporter fields.