NZLOAD is working while External Table in Netezza failing with count of bad input rows reached maxerrors

Anoop R picture Anoop R · Oct 7, 2015 · Viewed 14.9k times · Source

Intermittent issues with Netezza External Table .

External table is failing with the file generated by the system itself(Meant the file generated by external table not from other sources).But we tried to load the same file via nzload utility to another table and that is working without any issues.This issue is not consistent and unable to reproduce most of times.

CREATE EXTERNAL TABLE SP_PORTFOLIO_EXT_DATA_6128_140
(
    CLIENT_ID INTEGER,
    CONFIG_ID INTEGER,
    SCENARIO_ID INTEGER,
    PORTFOLIO_ID INTEGER,
    PORTFOLIO_NAME CHARACTER VARYING(200),
    CUSTOM13 CHARACTER VARYING(600),
    CUSTOM12 CHARACTER VARYING(500),
    CUSTOM11 CHARACTER VARYING(500),
    CUSTOM10 CHARACTER VARYING(500),
    CUSTOM9 CHARACTER VARYING(500),
    CUSTOM8 CHARACTER VARYING(500),
    CUSTOM7 CHARACTER VARYING(500),
    CUSTOM6 CHARACTER VARYING(2000),
    CUSTOM3 CHARACTER VARYING(500),
    CUSTOM2 CHARACTER VARYING(3000),
    CUSTOM1 CHARACTER VARYING(500),
    CREATIVE CHARACTER VARYING(512),
    PLACEMENT CHARACTER VARYING(5000),
    IMPRESSIONS NUMERIC(38,0),
    CLICKS NUMERIC(38,0),
    CONVERSIONS INTEGER,
    TRUE_CONVERSIONS NUMERIC(38,6),
    OPTMETRIC NUMERIC(38,6),
    LASTAD_OPTMETRIC NUMERIC(38,6),
    CURRSPEND NUMERIC(38,6)
)
USING
(
    DATAOBJECT('/san5/Netezza/CAR/CAR_ZEUS/SPBU/test/SP_PORTFOLIO_EXT_DATA_6128_140.csv')
    DELIMITER 254
    ESCAPECHAR '/'
    TIMESTYLE '24HOUR'
    LOGDIR '/tmp'
    Y2BASE 2000
    ENCODING 'internal'
);

The command completed successfully

select COUNT(*) from SP_PORTFOLIO_EXT_DATA_6128_140;
ERROR [HY000] ERROR:  External Table : count of bad input rows reached maxerrors limit

NZLOAD method

CREATE TABLE TEST_LOAD
(
    CLIENT_ID INTEGER,
    CONFIG_ID INTEGER,
    SCENARIO_ID INTEGER,
    PORTFOLIO_ID INTEGER,
    PORTFOLIO_NAME CHARACTER VARYING(200),
    CUSTOM13 CHARACTER VARYING(600),
    CUSTOM12 CHARACTER VARYING(500),
    CUSTOM11 CHARACTER VARYING(500),
    CUSTOM10 CHARACTER VARYING(500),
    CUSTOM9 CHARACTER VARYING(500),
    CUSTOM8 CHARACTER VARYING(500),
    CUSTOM7 CHARACTER VARYING(500),
    CUSTOM6 CHARACTER VARYING(2000),
    CUSTOM3 CHARACTER VARYING(500),
    CUSTOM2 CHARACTER VARYING(3000),
    CUSTOM1 CHARACTER VARYING(500),
    CREATIVE CHARACTER VARYING(512),
    PLACEMENT CHARACTER VARYING(5000),
    IMPRESSIONS NUMERIC(38,0),
    CLICKS NUMERIC(38,0),
    CONVERSIONS INTEGER,
    TRUE_CONVERSIONS NUMERIC(38,6),
    OPTMETRIC NUMERIC(38,6),
    LASTAD_OPTMETRIC NUMERIC(38,6),
    CURRSPEND NUMERIC(38,6)
)
DISTRIBUTE ON RANDOM;

# Loading data  from the same file using Nzload

nzload -host 10.200.29.30 -u xxxxx -pw xxxxx -db SPBU_REPORT_DB_TEST -t test_load -delim 254 -ctrlChars  -df /san5/Netezza/CAR/CAR_ZEUS/SPBU/test/SP_PORTFOLIO_EXT_DATA_6128_140.csv

Load session of table 'TEST_LOAD' completed successfully

[ja.prod@inet11026 ~]$ cat /san5/Netezza/CAR/CAR_ZEUS/SPBU/test/SP_PORTFOLIO_EXT_DATA_6128_140.csv|wc -l
191322

select count(*) from test_load;
191322

Adding the nzlog

File Buffer Size (MB): 8                  Load Replay Region (MB): 0
  Encoding:              INTERNAL           Max errors:            1
  Skip records:          0                  Max rows:              0
  FillRecord:            No                 Truncate String:       No
  Escape Char:           '/'                Accept Control Chars:  No
  Allow CR in string:    No                 Ignore Zero:           No
  Quoted data:           NO                 Require Quotes:        No

  BoolStyle:             1_0                Decimal Delimiter:     '.'

  Disable NFC:           No
  Date Style:            YMD                Date Delim:            '-'
  Y2Base:                2000
  Time Style:            24HOUR             Time Delim:            ':'
  Time extra zeros:      No

Found bad records

bad #: input row #(byte offset to last char examined) [field #, declaration] diagnostic, "text consumed"[last char examined]
----------------------------------------------------------------------------------------------------------------------------
1: 25(184) [21, INT4] expected field delimiter or end of record, "0"[.]

Statistics

  number of records read:      25
  number of bad records:       1
  -------------------------------------------------
  number of records loaded:    0

  Elapsed Time (sec): 0.0

-----------------------------------------------------------------------------
Load completed at: 08-Oct-15 09:59:04 EDT

The .nzbad data containing the bad row (with pipe symbols standing in for the actual delimiter for readbility):

140|1305|6128||NULL|SEO|SEO|test.com/vehicledetail/detail/632888199/overview|SEO|SEO|SEO|SEO Brand|SEO Brand|best Tracking|Google(Seo)|SEO|Impression Tracker|Unknown|0|1|0|0.000000|0.000000|0.000000|0.000000

Answer

ScottMcG picture ScottMcG · Oct 8, 2015

From the nzlog we can tell that the load is failing on the 25th row. Specifically, as it tries to load the 21st column it encounters a value that is not an integer.

The log shows that it encounters a 0 which is then followed by a period. So the data likely has something like 0.0 or 0.1234 which can't be loaded as an integer.

bad #: input row #(byte offset to last char examined) [field #, declaration] diagnostic, "text consumed"[last char examined]
----------------------------------------------------------------------------------------------------------------------------
1: 25(184) [21, INT4] expected field delimiter or end of record, "0"[.]

With the .nzbad data you have provided (here with '|' instead of your actual delimiter for readability purposes):

140|1305|6128||NULL|SEO|SEO|test.com/vehicledetail/detail/632888199/overview|SEO|SEO|SEO|SEO Brand|SEO Brand|best Tracking|Google(Seo)|SEO|Impression Tracker|Unknown|0|1|0|0.000000|0.000000|0.000000|0.000000

One thing I notice here is that you have a varchar field with '/' in it. One of the differences between your external table and your nzload approach is that the external table specifies escapechar '/' while the nzload doesn't.

You will find that your data 'test.com/vehicledetail/detail/632888199/overview' will be loaded as 'test.comvehicledetaildetail632888199overview' because the '/' characters will be removed as they are not themselves escaped (e.g. '//').

If a '/' directly preceded a column delimiter in the data, it will direct it to consider the column delimeter to be part of the data instead, and will think column 22 in the data is actually for column 21 in the table which would match what we are seeing here.