Uploaded image for project: 'IMPALA'
  1. IMPALA
  2. IMPALA-10267

Impala crashes in HdfsScanner::WriteTemplateTuples() with negative num_tuples

Attach filesAttach ScreenshotVotersWatch issueWatchersCreate sub-taskLinkCloneUpdate Comment AuthorReplace String in CommentUpdate Comment VisibilityDelete Comments
    XMLWordPrintableJSON

    Details

    • Epic Color:
      ghx-label-13

      Description

      An exhaustive job hit two Impalad crashes with the following stack:

       2  impalad!google::LogMessageFatal::~LogMessageFatal() + 0x9
          rbx = 0x0000000000000000   rbp = 0x00007f82f98ad7a0
          rsp = 0x00007f82f98ad6a0   r12 = 0x0000000000000000
          r13 = 0x00007f8306dd1690   r14 = 0x000000002f6631a0
          r15 = 0x0000000072b8f2f0   rip = 0x0000000005209129
          Found by: call frame info
       3  impalad!impala::HdfsScanner::WriteTemplateTuples(impala::TupleRow*, int) [hdfs-scanner.cc : 235 + 0xf]
          rbx = 0x0000000000000000   rbp = 0x00007f82f98ad7a0
          rsp = 0x00007f82f98ad6b0   r12 = 0x0000000000000000
          r13 = 0x00007f8306dd1690   r14 = 0x000000002f6631a0
          r15 = 0x0000000072b8f2f0   rip = 0x0000000002802013
          Found by: call frame info
       4  impalad!impala::HdfsAvroScanner::ProcessRange(impala::RowBatch*) [hdfs-avro-scanner.cc : 553 + 0x19]
          rbx = 0x0000000000000400   rbp = 0x00007f82f98adc60
          rsp = 0x00007f82f98ad7b0   r12 = 0x0000000000000000
          r13 = 0x00007f8306dd1690   r14 = 0x000000002f6631a0
          r15 = 0x0000000072b8f2f0   rip = 0x000000000283880d
          Found by: call frame info
       5  impalad!impala::BaseSequenceScanner::GetNextInternal(impala::RowBatch*) [base-sequence-scanner.cc : 189 + 0x2b]
          rbx = 0x0000000000000000   rbp = 0x00007f82f98adf40
          rsp = 0x00007f82f98adc70   r12 = 0x0000000000000000
          r13 = 0x00007f8306dd1690   r14 = 0x000000002f6631a0
          r15 = 0x0000000072b8f2f0   rip = 0x00000000029302b5
          Found by: call frame info
       6  impalad!impala::HdfsScanner::ProcessSplit() [hdfs-scanner.cc : 143 + 0x39]
          rbx = 0x000000000292fbd4   rbp = 0x00007f82f98ae000
          rsp = 0x00007f82f98adf50   r12 = 0x0000000000000000
          r13 = 0x00007f8306dd1690   r14 = 0x000000002f6631a0
          r15 = 0x0000000072b8f2f0   rip = 0x00000000028011c9
          Found by: call frame info
       7  impalad!impala::HdfsScanNode::ProcessSplit(std::vector<impala::FilterContext, std::allocator<impala::FilterContext> > const&, impala::MemPool*, impala::io::ScanRange*, long*) [hdfs-scan-node.cc : 500 + 0x28]
          rbx = 0x0000000000008000   rbp = 0x00007f82f98ae390
          rsp = 0x00007f82f98ae010   r12 = 0x0000000000000000
          r13 = 0x00007f8306dd1690   r14 = 0x000000002f6631a0
          r15 = 0x0000000072b8f2f0   rip = 0x000000000297aa3d
          Found by: call frame info
       8  impalad!impala::HdfsScanNode::ScannerThread(bool, long) [hdfs-scan-node.cc : 418 + 0x27]
          rbx = 0x00000001abc6a760   rbp = 0x00007f82f98ae750
          rsp = 0x00007f82f98ae3a0   r12 = 0x0000000000000000
          r13 = 0x00007f8306dd1690   r14 = 0x000000002f6631a0
          r15 = 0x0000000072b8f2f0   rip = 0x0000000002979dbe
          Found by: call frame info
       9  impalad!impala::HdfsScanNode::ThreadTokenAvailableCb(impala::ThreadResourcePool*)::{lambda()#1}::operator()() const + 0x30
          rbx = 0x0000000000000bbf   rbp = 0x00007f82f98ae770
          rsp = 0x00007f82f98ae760   r12 = 0x0000000008e18f40
          r13 = 0x00007f8306dd1690   r14 = 0x000000002f6631a0
          r15 = 0x0000000072b8f2f0   rip = 0x0000000002979126
          Found by: call frame info

      This seems to happen when running query_test/test_scanners_fuzz.py::TestScannersFuzzing::test_fuzz_alltypes on Avro. In reading the code in HdfsAvroScanner ProcessScanRanger(), it seems impossible for this value to be negative, so it's unclear what is happening.

        Attachments

        Issue Links

          Activity

            People

            • Assignee:
              sql_forever Qifan Chen
              Reporter:
              joemcdonnell Joe McDonnell

              Dates

              • Created:
                Updated:
                Resolved:

                Issue deployment