The " Unable to shutdown metastore client' error is coming from shutting down 
syncMetaStoreClient, as both syncMetaStoreClient and metaStoreClient shares the 
same client.  I think in Hive:: close method, we should not call 
syncMetaStoreClient.close().

On 7/9/18, 12:49 PM, "Zoltan Haindrich" <[email protected]> wrote:

    I've a feeling that sometimes the same issue happens in other tests - but I 
agree; disabling it will make our life easier - until the real cause is 
uncovered and fixed.
    
    cheers,
    Zoltan
    
    
    On 07/09/2018 09:05 AM, Deepak Jaiswal wrote:
    > Thanks Zoltan for the analysis. Perhaps we should disable the test in the 
meantime as it is blocking several people from committing.
    > 
    > I can go ahead and create a patch for it.
    > 
    > Regards,
    > Deepak
    > 
    > On 7/8/18, 11:33 PM, "Zoltan Haindrich" <[email protected]> wrote:
    > 
    >      Hello
    >      
    >      Thank you Deepak for taking a closer look! ....from what you've 
found I've noticed that the runtime of TestReplicationScenariosAcidTables have 
jumped up to ~2000sec in the
    >      runs which have failed....it seems like this problem is there for a 
long time now; I've found jira tickets in which this test was "timed out" and 
the HiveQA comment was
    >      date at April 03....so it's not entirely new...
    >      
    >      The problem which prohibits this test from completing successfully 
seems like that it has difficulties closing down the metastore client - which 
goes on for a while ...
    >      I don't know if this is an acid/replication/metastore/? issue...but 
it seems intermittent - I've a hunch that somehow it might happen more reliably 
with this test...I've
    >      opened HIVE-20121 to investigate this...
    >      
    >      2018-07-08T22:07:33,461 DEBUG [main] metastore.HiveMetaStoreClient: 
Unable to shutdown metastore client. Will try closing transport directly.
    >      org.apache.thrift.transport.TTransportException: Cannot write to 
null outputStream
    >      
    >      some links to more or less recent logs:
    >      
http://104.198.109.242/logs/PreCommit-HIVE-Build-12481/failed/240_UTBatch_itests__hive-unit_9_tests/maven-test.txt
    >      the hive.log is ~200M:
    >      
http://104.198.109.242/logs/PreCommit-HIVE-Build-12481/failed/240_UTBatch_itests__hive-unit_9_tests/logs/hive.log
    >      
    >      
    >      cheers,
    >      Zoltan
    >      
    >      On 07/08/2018 06:49 PM, Deepak Jaiswal wrote:
    >      > I am seeing tests timing out in my latest ptest run,
    >      >
    >      > https://builds.apache.org/job/PreCommit-HIVE-Build/12468/testReport
    >      > https://builds.apache.org/job/PreCommit-HIVE-Build/12468/console
    >      >
    >      > TestAlterTableMetadata - did not produce a TEST-*.xml file (likely 
timed out) (batchId=240)
    >      > TestAutoPurgeTables - did not produce a TEST-*.xml file (likely 
timed out) (batchId=240)
    >      > TestLocationQueries - did not produce a TEST-*.xml file (likely 
timed out) (batchId=240)
    >      > TestReplicationScenariosAcidTables - did not produce a TEST-*.xml 
file (likely timed out) (batchId=240)
    >      > TestSemanticAnalyzerHookLoading - did not produce a TEST-*.xml 
file (likely timed out) (batchId=240)
    >      > TestSparkStatistics - did not produce a TEST-*.xml file (likely 
timed out) (batchId=240)
    >      >
    >      >
    >      >  From the Hive QA homepage, the last stable build was 12444 
whereas the current run is 12473. I looked at some of the runs in between and 
it looks like most of the runs are failing due to the above batch of unit tests.
    >      >
    >      > Regards,
    >      > Deepak
    >      >
    >      
    >      
    > 
    
    

Reply via email to