Overview

Each time you run analysis on a model, PhixFlow creates a new set of data, called a recordset, for each non-transactional table. You can view this data by right-clicking on a table and selecting 

Recordsets remain available until you run an archive task on the table; see Task. 

From the System Console → Task Logs tab → Recordsets you can select an option to rollback or to repeat analysis where the analysis run   generated a number of cycles; see Understanding Cycles, below.

How To RollBack RecordSets

For non-transactional tables, which have a Period of Daily, Monthly or Variable, you can back-out the changes to data caused by an analysis run. This resets the data to a previous recordset.

To rollback to an earlier recordset, open the rollback window from one of the following places:

• In a model: right-click on a table icon and select 

  Insert excerpt
  _rollback _rollback nopanel true

   Table.
• In the repository: right-click on a table name and select   

  Insert excerpt
  _rollback _rollback nopanel true

   Table.
• In the repository: use Ctrl+click to select multiple tables, right-click one of the table names and select   

  Insert excerpt
  _rollback _rollback nopanel true

   Selected Objects.
• In a list of recordsets: select one or more recordsets, then right-click on one recordset and select

  Insert excerpt
  _rollback _rollback nopanel true

   Recordset. PhixFlow automatically populates the data for the rollback options, as it knows the properties of the selected recordset(s).
• In the System Console → Task Logs tab → Recordsets , select rollback.for a recordset.

PhixFlow opens the Rollback window where you can configure the rollback. 

During the rollback process, PhixFlow displays details in the console. For large data sets, it may take some time to rollback all the data in a recordset.

Anchor
options options

Rollback Window Options

In the Rollback window, select an option and specify the recordsets to rollback.

Example

When testing an analysis model it is likely you will read and clear all the data frequently, in this instance you would use the following configuration:

If you rollback data due to errors, you can repopulate an incomplete recordset with the correct data. This is because the all recordsets in a model that were created during the same analysis run have the same Run ID. PhixFlow can use the run ID to pull the correct data into an incomplete table.

To repopulate an incomplete recordset:

1. Go to the console.
2. Find the log messages related to rollback.
3. Right-click the log message and select Repeat to reinstate data; see Rollback or Repeat Cycles Window.

For information about rerunning incomplete recordsets in an transactional table; see the Managing Recordsets → Reprocess Recordset option.

Understanding Repeat Cycles 

Anchor
#cycles #cycles

Whenever an analysis model is run it will carry out a number of tasks e.g. read a number of files, generate a number of recordsets, export a number of files.

You can configure an analysis model so that in a single analysis session it actually runs several times. When a model is configured in this way, each run is called a cycle. All of the items processed or generated during that cycle (i.e. the files processed, recordsets created, files exported etc) are flagged with the cycle number. This cycle number is visible in the Log File for that analysis run on each of the relevant tabs showing files and stream sets processed.

The most common reason for configuring a model in this way is to allow a number of files to be processed in a single analysis session but to have the results from each file saved in a separate recordset. In this way, whenever a file fails to process, you can simply rollback the recordsets associated with that file rather than rolling back all of the recordsets associated with all of the files for the whole analysis session.

To configure a model to run multiple times in one analysis session:

1. At the start of the model, configure the file collector to process a single file at a time.
2. At the end of the model, from the final table create a push pipe that points back to the final table. 
3. In the push pipe, set positive offests.

When you run analysis on the model:

1. The file collector reads in one file.
2. PhixFlow then processes all tables in the model, generating a recordset for each table.
3. Each recordset, and any associated output file, is flagged with the cycle number 1.
4. When the last table in the model has finished producing its recordset, it sends a notification to itself down the push pipe.
5. Due to the positive offset in the push pipe, the notification is the signal to create another recordset, so the whole model runs again. This time, recordsets and other output files are flagged with cycle number 2.

In the log files you can see which items were processed in which cycle. For rolling back analysis runs, you can choose how many cycles to roll back. In the case where the last file processed was unsuccessful, you can choose to only roll back the last cycle.

From the System Console → Task Logs tab → Recordsets you can see the cycle number associated with each recordset. You have options to rollback data to repeat analysis for specific cycles.

Anchor
rerun rerun

Repopulating Incomplete Recordsets

If you rollback data due to errors, you can repopulate an incomplete recordset with the correct data. This is because the all recordsets in a model that were created during the same analysis run have the same Run ID. PhixFlow can use the run ID to pull the correct data into an incomplete table.

To repopulate an incomplete recordset:

1. Go to the console.
2. Find the log messages related to rollback.
3. Right-click the log message and in the Rollback window, select Repeat to reinstate data.

For information about rerunning incomplete recordsets in an transactional table; see the Managing Recordsets → Reprocess Recordset option.