marton-bod commented on a change in pull request #2228:
URL: https://github.com/apache/iceberg/pull/2228#discussion_r573154983
##########
File path:
mr/src/main/java/org/apache/iceberg/mr/hive/HiveIcebergOutputCommitter.java
##########
@@ -90,22 +93,34 @@ public void commitTask(TaskAttemptContext originalContext)
throws IOException {
TaskAttemptContext context =
TezUtil.enrichContextWithAttemptWrapper(originalContext);
TaskAttemptID attemptID = context.getTaskAttemptID();
- String fileForCommitLocation =
generateFileForCommitLocation(context.getJobConf(),
- attemptID.getJobID(), attemptID.getTaskID().getId());
- HiveIcebergRecordWriter writer =
HiveIcebergRecordWriter.getWriter(attemptID);
+ JobConf jobConf = context.getJobConf();
+ Map<String, HiveIcebergRecordWriter> writers =
HiveIcebergRecordWriter.getWriters(attemptID);
+ Map<String, String> outputs =
SerializationUtil.deserializeFromBase64(jobConf.get(InputFormatConfig.OUTPUT_TABLES));
- DataFile[] closedFiles;
- if (writer != null) {
- closedFiles = writer.dataFiles();
- } else {
- closedFiles = new DataFile[0];
+ ExecutorService tableExecutor = tableExecutor(jobConf, outputs.size());
+ try {
+ // Generates commit files for the target tables in parallel
+ Tasks.foreach(outputs.entrySet())
+ .stopOnFailure()
Review comment:
For `commitJob`, you're right, most FS problems would already be retried
when reading the forCommit files. For `commitTask`, I thought failing to write
the forCommit file due to some intermittent FS issue could be handled and
safely retried. Does it add too much to the complexity? I though at least for
commitTask it wouldn't, we would just use the retry call in the fluent API.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
