[ 
https://issues.apache.org/jira/browse/BEAM-12356?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17511287#comment-17511287
 ] 

Bruno Candido Volpato da Cunha commented on BEAM-12356:
-------------------------------------------------------

This error is reproducible on 2.37.0 as well.

However, it seems to be on the Reader side of the BigQueryIO, not the writer.

 

A huge leak of Gax-* threads can be seen, after running a Beam job using this 
source:
{code:java}
PCollection<TableRow> tableRows =
    pipeline.apply(
        "Read from BigQuery",
        BigQueryIO.readTableRows()
            .from(
                new TableReference()
                    .setProjectId(params.getTableProjectName())
                    .setDatasetId(params.getTableDatasetName())
                    .setTableId(params.getTableName()))
            .withMethod(Method.DIRECT_READ)); {code}
Apparently, BigQueryServicesImpl doesn't `.close()` the Bigquery instance 
`client`. Will try to get more details and post them here.

 

 

 

 

> BigQueryWriteClient in DatasetServiceImpl is not closed, which causes 
> "ManagedChannel allocation site" exceptions
> -----------------------------------------------------------------------------------------------------------------
>
>                 Key: BEAM-12356
>                 URL: https://issues.apache.org/jira/browse/BEAM-12356
>             Project: Beam
>          Issue Type: Bug
>          Components: io-java-gcp
>    Affects Versions: 2.29.0, 2.32.0, 2.33.0
>            Reporter: Minbo Bae
>            Assignee: Reuven Lax
>            Priority: P2
>             Fix For: 2.34.0
>
>         Attachments: bigquery_grpc.log
>
>          Time Spent: 5.5h
>  Remaining Estimate: 0h
>
> [BigQueryWriteClient|https://github.com/apache/beam/blob/v2.29.0/sdks/java/io/google-cloud-platform/src/main/java/org/apache/beam/sdk/io/gcp/bigquery/BigQueryServicesImpl.java#L461]
>  in DatasetServiceImpl (added at [https://github.com/apache/beam/pull/14309)] 
> is not closed.  This causes the error logs  in gRPC orphan channel clean up. 
> See "bigquery_grpc.log" in attachments which is extracted from GCP Dataflow. 
> I don't think this issue affect pipeline runs except the error logs, but 
> could you take a look at that?
> A similar issue is reported for {{CloudBigtableIO}} at 
> [https://github.com/googleapis/java-bigtable-hbase/issues/2658]
>  



--
This message was sent by Atlassian Jira
(v8.20.1#820001)

Reply via email to