...

Motivation

In data warehouse, it's a common requirement to query historical data. So we need to periodically merge incremental data with historical data to generate complete data of current time. For example, Hive 2.2 supports using MERGE INTO statement to merge source table's records into target table by performing update, insert and delete according to different filter, so a good way is taking historical table as target and incremental table as source, then using MERGE INTO to perform update (for existing records) and insert (for non-existing records). But when there exists a large amount of data, the merging may cost much time.

...

But in most scenarios, a table will generate too many snapshots, so Paimon lets users to configure snapshot expiration time to clean old, unused snapshots. Thus the snapshot of a specified time point may have expired when user want to query it. So we want to introduce savepoint for Paimon to persist a snapshot in a specified time point

To solve this problem, we propose to introduce a new mechanism `Tag`. A tag is created from snapshot and can keep longer. The purposes of tag are:

1. Fault recovery (or we can say disaster recovery). Users can ROLL BACK to a tag if needed. If user rollbacks to a tag, the table will hold the data in the tag and the data committed  after the tag will be deleted. In this scenario the tag is like backup, but more lightweight because the tag only record the metadata of data files instead of copying them.

2. Record versions of data at a longer interval (typically daily or weekly). With tag, user can query the old data in batch mode. In this scenario the tag is like long-lived snapshot.

Proposed Changes

Storage

Like snapshot, a new directory `/savepoint` tag` will be created under table directory for storing savepointstags. The qualified path for a savepoint tag file is `/path/to/table/savepointtag/savepointtag-<id>`, and the id is increased from 1.

New Classes

We need an entrance class `Savepoint` `Tag` to store the information of a savepointtag:

public class SavepointTag {
	/** Used for identify a savepointtag. */
	private final long id;

	/** Used to identify which snapshot does this tag mark. */
    private final long taggedSnapshotId;

   	/** Id of the current table schema of the records in this savepointtag. */
	private final long schemaId;

	/** The manifest list of all data of this savepointtag. */
	private final long fullManifestList;

	/** How many records in this savepointtag. */
	private final long recordCount;

	/** Getters. */
	
	/** Some util methods. */

	/** Return all {@link ManifestFileMeta} instances for full data manifests in this snapshottag. */
 	public List<ManifestFileMeta> dataManifests(ManifestList manifestList);
	
	/** Serialize to json. */
	public String toJson();

	/** Deserialize from json. */
	public static Savepoint fromJson(String json);

	/** Get a savepointtag from path. */
	public static SavepointTag fromPath(FileIO fileIO, Path path);
}

We need a `SavepointManager` `TagManager` to manage the savepoints tags (similar to `SnapshotManager`).

public class SavepointManagerTagManager {
	/** Return the root Directory of savepointstags. */
	public Path savepointDirectorytagDirectory();
 	
	/** Return the path of a savepointtag. */
	public Path savepointPathtagPath(long savepointIdtagId);

 	/** Get a savepointtag instance. */
	public SavepointTag savepointtag(long savepointIdtagId);

 	/** Check if a savepointtag exists. */
	public boolean savepointExiststagExists(long savepointIdtagId);
 	
	/** Return id of the earlistearliest savepointtag. */
 	public @Nullable Long earlistSavapointIdearliestTagId();

 	/** Return id of the latest savepointtag. */
    public @Nullable Long latestSavepointIdlatestTagId();
}

We need a `SavepointsTable``TagsTable`, which can provide information of savepoints tags as system table `#savepoints``#tags`.

The schema of TagsTable is in section `Public Interfaces`.

Usage

Periodically savepoints

Public Interfaces

Table Options 

Flink Actions

Now, 

Call Procedures (Future work)

1. Spark supports extension to extend SQL syntax. We can provide savepoint procedures after we support CALL statement for Spark.
2. Supporting CALL is in the road map of Flink. We can provide savepoint procedures after then. 

Proposed Changes

Storage

Like snapshot, a new directory `/savepoint` will be created under table directory for storing savepoints. The qualified path for a savepoint file is `/path/to/table/savepoint/savepoint-<id>`.

Creation & Deletion

Currently, we can provide two ways for user to control the creation and deletion of savepoints.

...

In the future, after Flink support stored procedure, we can provide procedures to managing savepoints.

Interaction with Snapshot

1. When creating savepoint, we will do a full-compact and pick the compacted snapshot to save.
2. When expiring snapshot, Paimon will check whether the data files are used by living snapshots. If not, they will be deleted. After we introduce savepoint, we should also check if the data files are used by savepoints. 
3. When deleting savepoint, we will check and delete unused datafiles (like we expire snapshot).

System Table

We suppose to introduce a system table `SavepointsTable`. The schema is:

savepoint_id BIGINT
schema_id BIGINT
save_time BIGINT
record_count BIGINT 

Usage

Time Travel to Savepoint (Batch Reading)

Currently, we support time traveling to a snapshot. Now we can support to travel to a savepoint.

...

For Flink, we can use dynamic option `scan.snapshot-id=<snapshot-id>`. We suppose to introduce a new core option `scan.savepoint-id` to support traveling to a savepoint.

Time Travel to Timestamp with Savepoint

Currently, we support time traveling to a time point. Paimon Scanner will find the snapshot of which commit time is closest to the specified time. The problem is all of the snapshots before the time point may have expired. So we suppose to try to find the closest savepoint in this scenario.

Compatibility, Deprecation, and Migration Plan

None.

Test Plan

UT tests: verify creating and reading savepoint

...

1. time travel to savepoint
2. expiration of snapshots won't delete data files pointed by savepoints

Rejected Alternatives

Support starting from a savepoint in streaming reading

Current design of savepoint just store full data manifests, so it's not able to support streaming reading now.