drill-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "abdelhakim deneche" <adene...@gmail.com>
Subject Re: Review Request 33903: DRILL-2878: FragmentExecutor.closeOutResources() is not called if an exception happens in the Foreman before the fragment executor starts running
Date Sun, 10 May 2015 15:39:20 GMT

This is an automatically generated e-mail. To reply, visit:

(Updated May 10, 2015, 3:39 p.m.)

Review request for drill, Chris Westin and Jacques Nadeau.


fixed WorkerBee.startFragmentPendingRemote() to ignore cancelled fragments
small fix in unit test

Bugs: DRILL-2878

Repository: drill-git



This is a "quick fix" that seem to solve the problem, at least for the cases I am able to
reproduce it for. closeOutResources() shouldn't throw any exception at this point because
we didn't even start running, and any allocation failures will be suppressed (do we want this

If this fix is acceptable I will go ahead and add a "private volatile boolean startedRunning"
that will be set to true in run() and used in cancel() to check if we need to call closeOutResources().

I will also add a unit test, I know how to reproduce the problem for both the root and intermediate
fragments, but I still need to find a proper way to detect that those fragments were not closed

Diffs (updated)

  exec/java-exec/src/main/java/org/apache/drill/exec/work/WorkManager.java a3ceb8f 
  exec/java-exec/src/main/java/org/apache/drill/exec/work/fragment/FragmentExecutor.java ddb828c


Diff: https://reviews.apache.org/r/33903/diff/


all unit tests are passing along with functional/tpch100

I will redo the tests once DRILL-2757 has been committed


abdelhakim deneche

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message