Github user DaveBirdsall commented on a diff in the pull request:
https://github.com/apache/incubator-trafodion/pull/1356#discussion_r158144904
--- Diff: docs/sql_reference/src/asciidoc/_chapters/sql_utilities.adoc ---
@@ -443,24 +443,39 @@ specify one or more of these options:
** `CONTINUE ON ERROR`
+
-LOAD statement will continue after errors encountered while scanning rows
from source table.
+LOAD statement will continue after ignorable errors while scanning rows
from source table or loading into the target table. The ignorable errors are
usually data conversion errors.
+
Errors during the load or sort phase will cause the LOAD statement to
abort.
+
-Error rows will be logged by default in HDFS files in the directory
`/user/trafodion/bulkload/logs`. The default name of the error files will be of
the form `ERR_<three-part-target-table-name>_<date>_<id>`, where `<id>` is a
numeric identifier unique to the process where the error was seen.
-+
-This option is implied if `LOG ERROR ROWS [TO _error-location-name_]` or
`STOP AFTER _num_ ERROR ROWS` is specified and it is not enabled by default.
+This option is implied if `LOG ERROR ROWS [TO _error-location-name_]` or
`STOP AFTER _num_ ERROR ROWS` is specified.
** `LOG ERROR ROWS [TO _error-location-name_]`
+*** Error rows
+
If error rows must be written to a specified location, then specify TO
_error-location-name_, otherwise they will be written to the default location.
+`_error-location-name_` must be a HDFS directory name to which trafodion
has write access.
+
-Error logs are written in separate files by the processes involved in the
load command under sub-directory representing the load command in the given
location.
-The actual log file location is displayed in the load command output.
+Error rows will be logged in HDFS files in the *directory*
`/user/trafodion/bulkload/logs` if the error log location is not specified.
++
+The default name of the *subdirectory* is
`_ERR_catalog.schema.target_table_date_id_`, where `_id_` is a numeric
identifier timestamp (YYYYMMDD_HHMMSS) unique to the process where the error
was seen.
++
+The default name of the *error file* is
`_loggingFileNamePrefix_catalog.schema.target_table_instanceID_`, where
`_loggingFileNamePrefix_` is hive_scan_err or traf_upsert_err depending on the
data source table, and `_instanceID_` is the ID of instance starting from 0,
generally there is only one instance.
--- End diff --
Suggest "...is the instance ID starting from 0, ..."
---
