sarutak commented on PR #58574:
URL: https://github.com/apache/spark/pull/58574#issuecomment-5582298588

   Thanks, @dongjoon-hyun. This all sounds good, and I'll go with (a) for local 
mode in this PR and file (b) (running `UserCredentialManager` in 
`LocalSchedulerBackend` for real Kerberos parity) as a follow-up.
   
   On the two constructor-window asks:
   
   - I'll have the selection phase log one line (INFO/WARN) noting that 
driver-side access to the wired scheme will fail until credentials are acquired 
at scheduler start, so the eventual executor-side `ClassNotFoundException` is 
easier to trace.
   - I'll document the `SparkSubmit` `FileSystem`-cache case as its own item, 
calling out that it affects the default K8s cluster-mode path (`--jars 
s3a://BUCKET/...` with output on the same bucket) and produces a mixed-identity 
job rather than a failure.
   
   The rest is as you listed: I/O-free selection without `init()`, WARN + 
idempotent fallback in `start()`, a single owner for the loader, `stop()` 
ordering, scaladoc/description updates, `SparkContext`-level tests, and moving 
the call earlier. I'll also fix the commit message to `SPARK-59296` on the next 
push. Will ping you when it's up.


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to