Github user steveloughran commented on a diff in the pull request:
https://github.com/apache/spark/pull/19404#discussion_r183482802
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/CompactibleFileStreamLog.scala
---
@@ -139,6 +139,9 @@ abstract class CompactibleFileStreamLog[T <: AnyRef :
ClassTag](
out.write('\n')
out.write(Serialization.write(data).getBytes(UTF_8))
}
+ if (out.isInstanceOf[FSDataOutputStream]) {
+ out.asInstanceOf[FSDataOutputStream].hflush
+ }
--- End diff --
Something like
When opening stream
```scala
var syncable: Option[Syncable]] = if (out.isInstanceOf[FSDataOutputStream])
Some(out.asInstanceOf[FSDataOutputStream]) else None
```
then when at flush time, any flush failure just downgrades to a "don't
flush any more"
```scala
try {
syncable.map(_.hflush)
} catch {
case _: Exception => syncable = None
}
```
I know, risky, but I'm not 100% sure that its always
UnsupportedOperationException which gets raised. If IO is failing, it'll
surface later on anyway.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
