…w forkjoinpool on each iteration

…sults & fix some potential test flakes) as suggested by @attilapiros during review (thanks)

…re-being-shutdown-flat

… can experiment with this.

…lso remove a thread sleep

… increasing the max allowed job time to try and avoid flakiness.

…Resolver.scala


add more information on unexpected block.

Co-authored-by: Attila Zsolt Piros <2017933+attilapiros@users.noreply.github.com>

…dex blocks have both been migrated, check that RDD blocks are duplicated not just broadcast blocks, make the number of partitions smaller so the test can run faster, avoid the Thread.sleep for all of the tests except for the midflight test where we need it, check for the broadcast blocks landing (further along in scheduling) beyond just task start, force fetching the shuffle block to local disk if in shuffle block test mode, start the job as soon as the first executor comes online.

…re-being-shutdown-flat

…py so lets call it baseRdd, and test both small blocksize to mem and not

… if we're testing shuffles or not

…ote: this fails in forced migrate to disk

…ote: this fails in forced migrate to disk + some logging

…e test now passes so yay), re-enable the other tests I disabled while debugging. Add a bit more logging.

…e timeouts.

…ed it

… logic in update block :)