Uploaded image for project: 'Phoenix'
  1. Phoenix
  2. PHOENIX-3521

Scan over local index may return incorrect result after flush & compaction

    XMLWordPrintableJSON

Details

    • Bug
    • Status: Open
    • Major
    • Resolution: Unresolved
    • 4.7.0, 4.8.0
    • None
    • None
    • None

    Description

      Following code can be used to reproduce:

      @Test
        public void testit() throws Exception {
          Connection conn = DriverManager.getConnection("jdbc:phoenix:localhost");
          final PhoenixConnection phxConn = conn.unwrap(PhoenixConnection.class);
          while(true) {
            String t2 = "DROP TABLE IF EXISTS GIGANTIC_TABLE";
            conn.createStatement().execute(t2);
            t2 = "CREATE TABLE IF NOT EXISTS GIGANTIC_TABLE (ID INTEGER PRIMARY KEY, unsig_id UNSIGNED_INT, big_id BIGINT, unsig_long_id UNSIGNED_LONG, tiny_id TINYINT, unsig_tiny_id UNSIGNED_TINYINT, small_id SMALLINT, unsig_small_id UNSIGNED_SMALLINT, float_id FLOAT, unsig_float_id UNSIGNED_FLOAT, double_id DOUBLE, unsig_double_id UNSIGNED_DOUBLE, decimal_id DECIMAL, boolean_id BOOLEAN, time_id TIME, date_id DATE, timestamp_id TIMESTAMP, unsig_time_id TIME, unsig_date_id DATE, unsig_timestamp_id TIMESTAMP, varchar_id VARCHAR (30), char_id CHAR (30), binary_id VARCHAR (100), varbinary_id VARCHAR (100), array_id VARCHAR[])";
            conn.createStatement().execute(t2);
            CsvBulkLoadTool csvBulkLoadTool = new CsvBulkLoadTool();
            csvBulkLoadTool.setConf(new Configuration());
            int exitCode = csvBulkLoadTool.run(new String[]{
                    "--input", "/tmp/1.csv",
                    "--table", "GIGANTIC_TABLE",
                    "-d", ",",
                    "-a", ";",
                    "-q", "\"\"\""
            });
            assertEquals(0, exitCode);
              for(int j = 0; j < 5; j++) {
                  t2 = "DROP INDEX IF EXISTS LOCAL_INDEX_COLUMN_TYPE_char_id on GIGANTIC_TABLE";
                  phxConn.createStatement().execute(t2);
                  t2 = "CREATE LOCAL INDEX LOCAL_INDEX_COLUMN_TYPE_char_id ON GIGANTIC_TABLE (char_id)";
                  phxConn.createStatement().execute(t2);
                  String query = "SELECT count(*) FROM GIGANTIC_TABLE WHERE char_id like '%a%'";
                  for (int i = 0; i < 5 ; i++) {
                      ResultSet rs = phxConn.createStatement().executeQuery(query);
                      while (rs.next()) {
                          int result = rs.getInt(1);
                          if (result == 500000) {
                              LOG.error("OK");
                          } else {
                              LOG.error("FAILURE!!!!");
                          }
                      }
                  }
              }
          }
        }
      

      The issue is quite hard to reproduce and sometimes it requires several/many hours to get it. When it fails the result set contains value higher than 500k.

      Attachments

        1. gen.py
          2 kB
          Sergey Soldatov

        Activity

          People

            sergey.soldatov Sergey Soldatov
            sergey.soldatov Sergey Soldatov
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

            Dates

              Created:
              Updated: