[HIVE-10384] RetryingMetaStoreClient does not retry wrapped TTransportExceptions

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.2.0
• Component/s: Clients
• Assignee: Chaoyu Tang Assign to me
• Reporter: Eric Liang
• Created: 17/Apr/15 21:37
• Updated: 27/Jun/17 00:06
• Resolved: 30/Apr/15 18:15

Description

This bug is very similar to HIVE-9436, in that a TTransportException wrapped in a MetaException will not be retried. RetryingMetaStoreClient has a block of code above the MetaException handler that retries thrift exceptions, but this doesn't work when the exception is wrapped.

        if ((e.getCause() instanceof TApplicationException) ||
            (e.getCause() instanceof TProtocolException) ||
            (e.getCause() instanceof TTransportException)) {
          caughtException = (TException) e.getCause();
        } else if ((e.getCause() instanceof MetaException) &&
            e.getCause().getMessage().matches("(?s).*JDO[a-zA-Z]*Exception.*")) {
          caughtException = (MetaException) e.getCause();

Issue Links

• is related to: HIVE-9436 RetryingMetaStoreClient does not retry JDOExceptions
• is related to: HIVE-10608 Fix useless 'if' statement in RetryingMetaStoreClient (135)
• relates to: HIVE-9642 Hive metastore client retries don't happen consistently for all api calls
• relates to: HIVE-16966 Remove String Checks For Error Handling