[
https://issues.apache.org/jira/browse/HIVE-2444?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16029226#comment-16029226
]
Peter Vary commented on HIVE-2444:
----------------------------------
I think it is solved by:
- HIVE-8829 - Upgrade to Thrift 0.9.2
Thrift 0.9.2 includes:
- THRIFT-2788 - Java client and server does not enable tcp keep alive., which
enables keepalive
> Add TCP keepalive option for the hive jdbc driver
> -------------------------------------------------
>
> Key: HIVE-2444
> URL: https://issues.apache.org/jira/browse/HIVE-2444
> Project: Hive
> Issue Type: Improvement
> Components: JDBC
> Affects Versions: 0.7.1
> Reporter: Bob Tiernay
>
> When issuing long running jdbc queries it is quite plausibe that due to
> stateful firewall rules connections may be dropped due to inactivity. Having
> the ability specify this as a jdbc property would help alleviate this issue
> when firewall connection inactivity timeouts are greater than the tcp
> keepalive time (default of 2 hrs).
> As a temporary workaround, one can use reflection on the jdbc connection to
> set this option:
> {code}
> public static Connection getHiveConnection() {
> Connection hiveConnection = // create connection
> try {
> Socket socket = getConnectionSocket( hiveConnection );
> socket.setKeepAlive( true );
> } catch( SocketException e ) {
> throw new RuntimeException( e );
> }
> return hiveConnection;
> }
> private static Socket getConnectionSocket( Connection connection ) {
> final Field fields[] = connection.getClass().getDeclaredFields();
> for( int i = 0; i < fields.length; ++i ) {
> if( "transport".equals( fields[i].getName() ) ) {
> try {
> fields[i].setAccessible( true );
> TSocket transport = (TSocket) fields[i].get( connection );
> return transport.getSocket();
> } catch( Exception e ) {
> throw new RuntimeException( e );
> }
> }
> }
> return null;
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
