Data Management

This tutorial is going to introduce you to the conceptual details of data management like:

Loading Data

  • Scenario

    After creating a table, you can load data to the table using the LOAD DATA command. The loaded data is available for querying. When data load is triggered, the data is encoded in CarbonData format and copied into HDFS CarbonData store path (specified in file) in compressed, multi dimensional columnar format for quick analysis queries. The same command can be used to load new data or to update the existing data. Only one data load can be triggered for one table. The high cardinality columns of the dictionary encoding are automatically recognized and these columns will not be used for dictionary encoding.

  • Procedure

    Data loading is a process that involves execution of multiple steps to read, sort and encode the data in CarbonData store format. Each step is executed on different threads. After data loading process is complete, the status (success/partial success) is updated to CarbonData store metadata. The table below lists the possible load status.

Status Description
Success All the data is loaded into table and no bad records found.
Partial Success Data is loaded into table and bad records are found. Bad records are stored at carbon.badrecords.location.

In case of failure, the error will be logged in error log. Details of loads can be seen with SHOW SEGMENTS command. The show segment command output consists of :

  • SegmentSequenceId
  • Status
  • Load Start Time
  • Load End Time

The latest load will be displayed first in the output.

Refer to DML operations on CarbonData for load commands.

Deleting Data

  • Scenario

    If you have loaded wrong data into the table, or too many bad records are present and you want to modify and reload the data, you can delete required data loads. The load can be deleted using the Segment Sequence Id or if the table contains date field then the data can be deleted using the date field. If there are some specific records that need to be deleted based on some filter condition(s) we can delete by records.

  • Procedure

    The loaded data can be deleted in the following ways:

    • Delete by Segment ID

      After you get the segment ID of the segment that you want to delete, execute the delete command for the selected segment. The status of deleted segment is updated to Marked for delete / Marked for Update.

SegmentSequenceId Status Load Start Time Load End Time
0 Success 2015-11-19 19:14:... 2015-11-19 19:14:...
1 Marked for Update 2015-11-19 19:54:... 2015-11-19 20:08:...
2 Marked for Delete 2015-11-19 20:25:... 2015-11-19 20:49:...
  • Delete by Date Field

    If the table contains date field, you can delete the data based on a specific date.

  • Delete by Record

    To delete records from CarbonData table based on some filter Condition(s).

    For delete commands refer to DML operations on CarbonData.

  • NOTE:

    • When the delete segment DML is called, segment will not be deleted physically from the file system. Instead the segment status will be marked as "Marked for Delete". For the query execution, this deleted segment will be excluded.

    • The deleted segment will be deleted physically during the next load operation and only after the maximum query execution time configured using "max.query.execution.time". By default it is 60 minutes.

    • If the user wants to force delete the segment physically then he can use CLEAN FILES Command.

Example :


This DML will physically delete the segment which are "Marked for delete" and "Compacted" immediately.

Compacting Data

  • Scenario

    Frequent data ingestion results in several fragmented CarbonData files in the store directory. Since data is sorted only within each load, the indices perform only within each load. This means that there will be one index for each load and as number of data load increases, the number of indices also increases. As each index works only on one load, the performance of indices is reduced. CarbonData provides provision for compacting the loads. Compaction process combines several segments into one large segment by merge sorting the data from across the segments.

  • Procedure

    There are two types of compaction Minor and Major compaction.

    • Minor Compaction

      In minor compaction the user can specify how many loads to be merged. Minor compaction triggers for every data load if the parameter is set. If any segments are available to be merged, then compaction will run parallel with data load. There are 2 levels in minor compaction.

      • Level 1: Merging of the segments which are not yet compacted.
      • Level 2: Merging of the compacted segments again to form a bigger segment.
    • Major Compaction

      In Major compaction, many segments can be merged into one big segment. User will specify the compaction size until which segments can be merged. Major compaction is usually done during the off-peak time.

    There are number of parameters related to Compaction that can be set in file

Parameter Default Application Description Valid Values
carbon.compaction.level.threshold 4, 3 Minor This property is for minor compaction which decides how many segments to be merged. Example: If it is set as "2, 3", then minor compaction will be triggered for every 2 segments in level 1. 3 is the number of level 1 compacted segment which is further compacted to new segment in level 2. NA
carbon.major.compaction.size 1024 MB Major Major compaction size can be configured using this parameter. Sum of the segments which is below this threshold will be merged. NA
carbon.numberof.preserve.segments 0 Minor/Major This property configures number of segments to preserve from being compacted. Example: carbon.numberof.preserve.segments=2 then 2 latest segments will always be excluded from the compaction. No segments will be preserved by default. 0-100
carbon.allowed.compaction.days 0 Minor/Major Compaction will merge the segments which are loaded within the specific number of days configured. Example: If the configuration is 2, then the segments which are loaded in the time frame of 2 days only will get merged. Segments which are loaded 2 days apart will not be merged. This is disabled by default. 0-100
carbon.number.of.cores.while.compacting 2 Minor/Major Number of cores which is used to write data during compaction. 0-100

For compaction commands refer to DDL operations on CarbonData

Updating Data

  • Scenario

    Sometimes after the data has been ingested into the System, it is required to be updated. Also there may be situations where some specific columns need to be updated on the basis of column expression and optional filter conditions.

  • Procedure

    To update we need to specify the column expression with an optional filter condition(s).

    For update commands refer to DML operations on CarbonData.