(2008) Point-In-Time Recovery for HANA

Owned by SNP
Aug 03, 2020
2 min read

The PIT recovery feature allows to make the state of a system consistent. This functionality allows to synchronize HANA with the archive. Perhaps the archive has been rolled back to a certain moment and HANA is now in an inconsistent state. This functionality analyzes the situation and chooses a timestamp to perform the rollback on HANA.

Before you can start to use these functionalities, you should enable Point-In-Time Recovery (PIT) and set a binary storage in DataTiering Settings Tab.


Using the Task Editor

To synchronize the Hadoop with the primary database:

  • Launch transaction /DVD/TH_EDIT, open the process /DVD/SM_HANA_CHL_CON and hit F8.

       

  •  Create a new task with a new Run ID by clicking the button below (or by hitting F8 key).


       

  • Name your Run ID.

         

  • Now you are ready to run your PIT tasks.

       

If you look at the task hierarchy on the screen, you will see that the PIT recovery process contains 6 different subtasks divided in 2 groups:

  • Analysis: This optional subtask analyzes what operations are required to make the archive consistent with the primary database.
  • Execution: This subtask performs operations to make the archive consistent with the primary database.

For each subgroup there are three operations:

  • Generate the analysis / synchronization tasks: You can insert the DataProviders you are testing as a parameter here and then execute a report which generates a task for each DataProvider.
  • Analyze the DataProviders / Execute the synchronization process: Executes either the analysis tasks or the synchronization tasks that were previously generated on the DataProviders.
  • Show the results of the analysis / synchronization: Displays a report detailing the operations that are completed on each DataProvider.



To delete data used for PIT recovery older than a certain timestamp, use report /DVD/SM_HANA_CHL_DELETE. The idea is schedule the execution of this job periodically to avoid growth of the database.


To clean the details of these tasks there is the report /DVD/SM_HANA_CHL_CLEAN_RUN which takes as input a set of Run IDs and cleans all information about them.

If for some reason you want to re-open and existing Run ID you can do it via transaction /DVD/TH_MONITOR and entering your Run ID (in our example "PIT_RUN") in the Task run ID field.


Metadata synchronization

DataTiering metadata (e.g. requests, statistics) are also synchronized during the synchronization process.

When request is rollbacked to the point in time where data were successfully offloaded, phase of the request is set to 'Offloaded' and it is possible to reload this request again.

When request is rollbacked to the point where it did not exist yet, phase of the request is set to 'Rollbacked' and 'Lines Read' to 0:

Additional information about the rollback can be found in request's logs and statistics.