[SPARK-48056][CONNECT][FOLLOW-UP] Scala Client re-execute plan if a SESSION_NOT_FOUND error is raised and no partial response was received #46971
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…error is raised and no partial response was received
ghost
mentioned this pull request
added 2 commits
June 13, 2024 13:12
juliuszsompolski
approved these changes
Jun 13, 2024
.../client/jvm/src/test/scala/org/apache/spark/sql/connect/client/SparkConnectClientSuite.scala
Outdated
Show resolved
Hide resolved
added 2 commits
June 13, 2024 13:16
Author
|
@HyukjinKwon Hi! Could you review and possibly merge this PR? It is a follow-up change of #46297 which was merged in May. |
zhengruifeng
approved these changes
Jun 14, 2024
ghost
changed the title
zhengruifeng
changed the title
ghost
deleted the
5 tasks
vkorukanti
pushed a commit
to delta-io/delta
that referenced
this pull request
Aug 21, 2024
…e available in Delta Connect testing (#3576) ## Description For local E2E Delta Connect testing, we also designed an [util class](https://github.com/delta-io/delta/blob/01bf60743b77c47147843e9083129320490f1629/spark-connect/client/src/test/scala-spark-master/io/delta/connect/tables/RemoteSparkSession.scala#L62) to start a local server in a different process similar to [SparkConnect](https://github.com/apache/spark/blob/ba208b9ca99990fa329c36b28d0aa2a5f4d0a77e/connector/connect/client/jvm/src/test/scala/org/apache/spark/sql/test/RemoteSparkSession.scala#L37). We noticed that the server takes a random amount of seconds to start up, and back then we received the error `INVALID_HANDLE.SESSION_NOT_FOUND] The handle 746e6c86-9fa9-4b08-9572-388c20eaed47 is invalid. Session not found. SQLSTATE: HY000"`, so what we did is to add a 10s `Thread.sleep` before starting the client. This is not robust, so we are removing the `Thread.sleep`. This should work because: 1. The SparkSession's builder here already uses the default [Configuration](https://github.com/apache/spark/blob/3edc9c23a723a92c5a951cea0436529de65c640a/connector/connect/client/jvm/src/main/scala/org/apache/spark/sql/SparkSession.scala#L891) of the `SparkConnectClient` which includes a default retry policy. 2. Spark patches the error `INVALID_HANDLE.SESSION_NOT_FOUND` in this [PR](apache/spark#46971) at some point, so we should be able to retry even if encountering this error. ## How was this patch tested? Existing UTs.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What changes were proposed in this pull request?
This change lets a Scala Spark Connect client reattempt execution of a plan when it receives a SESSION_NOT_FOUND error from the Spark Connect service if it has not received any partial responses.
This is a Scala version of the previous fix of the same issue - #46297.
Why are the changes needed?
Spark Connect clients often get a spurious error from the Spark Connect service if the service is busy or the network is congested. This error leads to a situation where the client immediately attempts to reattach without the service being aware of the client; this leads to a query failure.
Does this PR introduce any user-facing change?
Prevoiusly, a Scala Spark Connect client would fail with the error code "INVALID_HANDLE.SESSION_NOT_FOUND" in the very first attempt to make a request to the service, but with this change, the client will automatically retry.
How was this patch tested?
Attached unit test.
Was this patch authored or co-authored using generative AI tooling?
No.