[openhouse] Implement TableOperations-specific FileIO instance #241
Conversation
| @@ -88,6 +92,8 @@ public class OpenHouseCatalog extends BaseMetastoreCatalog | |||
|
|
|||
| protected Map<String, String> properties; | |||
|
|
|||
| private Cache<TableOperations, FileIO> fileIOCloser; | |||
There was a problem hiding this comment.
these look like temporary duplication of code that will be available in upstream iceberg with version bump. Can we add a todo identifiying the PR and version of iceberg that introduced the fileIOcloser and instructions on removing the code duplication later?
| .snapshotApi(snapshotApi) | ||
| .cluster(cluster) | ||
| .build(); | ||
| FileIO tableOperationsLocalFileIO = loadFileIO(properties); |
There was a problem hiding this comment.
can we also change the signature of loadFileIO to following:
protected FileIO loadFileIO(TableIdentifier tableIdentifier, Map<String, String> properties) {
reason:
Each table can have its own storage, table1 -> hdfs, table2 -> s3, table3 -> adls . We'd need to instantiate/load the right FileIO per table and properties.
| @@ -592,4 +603,25 @@ private TableMetadata createStagedMetadata() { | |||
| return new StaticTableOperations(tableLocation, fileIO).refresh(); | |||
| } | |||
| } | |||
|
|
|||
| @Override | |||
| public void close() { | |||
There was a problem hiding this comment.
can we add unit tests to validate this logic?
| @@ -41,6 +42,7 @@ dependencies { | |||
| exclude group: 'com.zaxxer', module: 'HikariCP-java7' | |||
| exclude group: 'org.apache.commons', module: 'commons-lang3' | |||
| } | |||
| implementation("com.github.ben-manes.caffeine:caffeine:" + caffeineVersion) | |||
| } | |||
There was a problem hiding this comment.
can we also do output jar check to see if the newly added library is appropriately relocated? the output jar should look as follows:
| @@ -2,6 +2,9 @@ | |||
|
|
|||
| import static com.linkedin.openhouse.javaclient.OpenHouseTableOperations.*; | |||
There was a problem hiding this comment.
for fileio per table, we might also need similar changes in OpenHouseInternalCatalog https://github.com/linkedin/openhouse/blob/main/iceberg/openhouse/internalcatalog/src/main/java/com/linkedin/openhouse/internal/catalog/OpenHouseInternalCatalog.java
| @@ -88,6 +92,8 @@ public class OpenHouseCatalog extends BaseMetastoreCatalog | |||
|
|
|||
| protected Map<String, String> properties; | |||
|
|
|||
| private Cache<TableOperations, FileIO> fileIOCloser; | |||
There was a problem hiding this comment.
Just curious. How long the entries are present in the cache? Wondering if the cache will be populated with too many objects if there are more tables operation done in a short period of time. Do you think is going to consume more memory?
| private Cache<TableOperations, FileIO> newFileIOCloser() { | ||
| return Caffeine.newBuilder() | ||
| .weakKeys() | ||
| .removalListener( |
There was a problem hiding this comment.
When the listener is invoked to perform close?
Summary
[Issue][(https://github.com/linkedin/openhouse/issues/#nnn)] Briefly discuss the summary of the changes made in this
pull request in 2-3 lines.](#240)
In order to implement FileIOs that are TableOperations-specific, creating a FileIO instance on a per TableOperations-basis is required.
This is similar to how other Iceberg-compliant Catalogs are implemented: apache/iceberg#10893 and implemented in the same way (taken from the HiveCatalog / GlueCatalog implementations before it got refactored into FileIOTracker)
This feature is needed to implement data access tokens.
Changes
For all the boxes checked, please include additional details of the changes made in this pull request.
Testing Done
TODO
For all the boxes checked, include a detailed description of the testing done for the changes made in this pull request.
Additional Information
For all the boxes checked, include additional details of the changes made in this pull request.