File Format and Examples

If a file is used (-f option) to specify the tables to purge, the format of the file specified in the above point 4-b is as follows:

Each line in the text file contains five fields that are TAB separated.

The fields are:

 tablename<TAB>columnname<TAB>additional key names<TAB>pk value<TAB>prk value 

The following are the definitions of these fields.

1. tablename : The name of the table from which data will be removed if it satisfies the criteria.
2. columnname : The name of the column that holds the timestamp.
3. additional key names : The name(s) of other columns that are part of the primary key until the column that holds timestamp. Each column name should be comma separated.
4. pk value : Whether or not the column that holds timestamp is part of the partition key. 0 indicates no and 1 indicates yes.
5. prk value : Whether or not the column that holds timestamp is a part of the Primary key. 0 indicates no and 1 indicates yes.

Important

A file table_specs.txt is included with the JAR that contains specifications for all the Workbench tables. This can be used to remove data from all Workbench tables.

Some Examples

Example 1

Consider the following table definition:

   CREATE TABLE act.alerts_triggered_by_call
            call_id uuid,
            alert_id uuid,
            alert_condition text,
            alert_timestamp timestamp,
            alert_type text,
            parameters map<text, text>,
            PRIMARY KEY (call_id, alert_id)
   ) WITH CLUSTERING ORDER BY (alert_id ASC)

The entry in the file for this would be:

   act.alerts_triggered_by_call	    alert_timestamp	  call_id, alert_id        no	no 

Since alert_timestamp is not part of the partition key, it is not a part of the primary key.

A sample of the command for this table would be:

   java -jar DataPurgeUtility.jar -h 135.17.180.184 -p 9042 -d 30 -t act.alerts_triggered_by_call -c 
   alert_timestamp -a "call_id,alert_id" -pk 0 -prk 0

Example 2

Consider the following table:

   CREATE TABLE beholder.changes (
          timemodified timestamp,
          dbid int,
          changekey text,
          changeby text,
          changetopobjectname text,
          changevalue text,
          objecthostip text,
          objecthostname text,
         operationtype text,
         PRIMARY KEY (timemodified, dbid, changekey)
   ) WITH CLUSTERING ORDER BY (dbid ASC, changekey ASC)

The entry in the file for this would be:

   beholder.changes	      timemodified	    dbid,changekey	  yes	yes

Since timemodified is a part of the partition key and it is a part of the primary key.

A sample of the command for this table would be:

   java -jar DataPurgeUtility.jar -h 135.17.180.184 -p 9042 -d 30 -t beholder.changes -c timemodified -a "dbid,changekey" -pk 1 -prk 1