Joe McDonnell created IMPALA-10058: -------------------------------------- Summary: Kudu queries hit error "Unable to deserialize scan token" Key: IMPALA-10058 URL: https://issues.apache.org/jira/browse/IMPALA-10058 Project: IMPALA Issue Type: Bug Components: Backend Affects Versions: Impala 4.0 Reporter: Joe McDonnell
I have seen a few test runs fail with a large number of Kudu tests failing with: {noformat} ImpalaBeeswaxException: ImpalaBeeswaxException: Query aborted:Unable to deserialize scan token for node with id '0' for Kudu table 'impala::functional_kudu.alltypestiny': Not found: the table does not exist: table_name: ""{noformat} In the Impalad log, the errors looks like this: {noformat} I0804 18:35:19.075631 18788 status.cc:129] 1c4fdf1a9de8d577:32d91ed500000002] Unable to deserialize scan token for node with id '1' for Kudu table 'impala::functional_kudu.alltypessmall': Not found: the table does not exist: table_name: "" @ 0x1cabde1 impala::Status::Status() @ 0x28e97a9 impala::KuduScanner::OpenNextScanToken() @ 0x284eb33 impala::KuduScanNode::ProcessScanToken() @ 0x284f15e impala::KuduScanNode::RunScannerThread() @ 0x284e351 _ZZN6impala12KuduScanNode17ThreadAvailableCbEPNS_18ThreadResourcePoolEENKUlvE_clEv @ 0x284fa48 _ZN5boost6detail8function26void_function_obj_invoker0IZN6impala12KuduScanNode17ThreadAvailableCbEPNS3_18ThreadResourcePoolEEUlvE_vE6invokeERNS1_15function_bufferE @ 0x2062c1d boost::function0<>::operator()() @ 0x26867cf impala::Thread::SuperviseThread() @ 0x268e76c boost::_bi::list5<>::operator()<>() @ 0x268e690 boost::_bi::bind_t<>::operator()() @ 0x268e651 boost::detail::thread_data<>::run() @ 0x3e616a1 thread_proxy @ 0x7fe872ef4e24 start_thread @ 0x7fe86f9c934c __clone{noformat} This error would be coming from the Kudu client in this code in KuduScanner: {noformat} kudu::client::KuduScanner* scanner; KUDU_RETURN_IF_ERROR(kudu::client::KuduScanToken::DeserializeIntoScanner( scan_node_->kudu_client(), scan_token, &scanner), BuildErrorString("Unable to deserialize scan token"));{noformat} This has happened multiple times in the docker-based tests, but I have also seen a couple jobs with the normal test runs. -- This message was sent by Atlassian Jira (v8.3.4#803005)