Github user mgaido91 commented on a diff in the pull request:
https://github.com/apache/spark/pull/18329#discussion_r123445514
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/streaming/DataStreamWriter.scala
---
@@ -264,12 +281,12 @@ final class DataStreamWriter[T] private[sql](ds:
Dataset[T]) {
df,
sink,
outputMode,
- useTempCheckpointLocation = true,
+ useTempCheckpointLocation = isTempCheckpointLocationAvailable,
trigger = trigger)
} else {
val (useTempCheckpointLocation, recoverFromCheckpointLocation) =
if (source == "console") {
- (true, false)
+ (isTempCheckpointLocationAvailable, false)
--- End diff --
Well, actually I don't think "java.io.tmpdir" will ever be on a filesystem
different from the local one. But, the other PR forces the metadata to be
written on the local filesystem, despite the default one is different (for
instance it can be HDFS).
This means that in a distributed environment, which should be fault
tolerant, with that patch if a node fails we loose the metadata. Since one of
the involved sink is the `foreach` one, which can be used to write the data
somewhere (for example HBase or Kafka), I think that forcing the user to
specify a `checkpointLocation` which is created on the `defaultFs` (in this
case HDFS) would be a better option.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
