Oh that looks like a considerable bug and could explain the not so good performance of the Portable VR tests (if this is happening for every call as you mention). Looks like more worth of a JIRA even than the FlinkRequiresStableInputTest.
On Mon, Mar 23, 2020 at 7:42 PM Reuven Lax <[email protected]> wrote: > > It looks like every call to getParDoPayload tries to zip up the entire > staging directory. This is pretty bad, as we call getParDoPayload in many > places, and most of those places don't expect it to have such a side effect. > > On Mon, Mar 23, 2020 at 11:38 AM Reuven Lax <[email protected]> wrote: >> >> I can file a JIRA, but does anyone know why this test is so flaky? Do we >> simply need to give it longer to run? >> >> https://builds.apache.org/job/beam_PreCommit_Java_Phrase/1912/testReport/junit/org.apache.beam.runners.flink/FlinkRequiresStableInputTest/testParDoRequiresStableInput/
