Package org.apache.cassandra.repair
Interface TableRepairManager
- All Known Implementing Classes:
CassandraTableRepairManager
public interface TableRepairManager
Table level hook for repair
-
Method Summary
Modifier and TypeMethodDescriptiongetValidationIterator(Collection<Range<Token>> ranges, TimeUUID parentId, TimeUUID sessionID, boolean isIncremental, long nowInSec, TopPartitionTracker.Collector topPartitionCollector) Return a validation iterator for the given parameters.voidincrementalSessionCompleted(TimeUUID sessionID) Called when the given incremental session has completed.voidsnapshot(String name, Collection<Range<Token>> ranges, boolean force) For snapshot repairs.Future<?>submitValidation(Callable<Object> validation) Begin execution of the given validation callable.
-
Method Details
-
getValidationIterator
ValidationPartitionIterator getValidationIterator(Collection<Range<Token>> ranges, TimeUUID parentId, TimeUUID sessionID, boolean isIncremental, long nowInSec, TopPartitionTracker.Collector topPartitionCollector) throws IOException, NoSuchRepairSessionException Return a validation iterator for the given parameters. If isIncremental is true, the iterator must only include data previously isolated for repair with the given parentId. nowInSec should determine whether tombstones should be purged or not. -
submitValidation
Begin execution of the given validation callable. Which thread pool a validation should run in is an implementation detail. -
incrementalSessionCompleted
Called when the given incremental session has completed. Because of race and failure conditions, implementors should not rely only on receiving calls from this method to determine when a session has ended. Implementors can determine if a session has finished by calling ActiveRepairService.instance.consistent.local.isSessionInProgress. Just because a session has completed doesn't mean it's completed succesfully. So implementors need to consult the repair service at ActiveRepairService.instance.consistent.local.getFinalSessionRepairedAt to get the repairedAt time. If the repairedAt time is zero, the data for the given session should be demoted back to unrepaired. Otherwise, it should be promoted to repaired with the given repaired time. -
snapshot
For snapshot repairs. A snapshot of the current data for the given ranges should be taken with the given name. If force is true, a snapshot should be taken even if one already exists with that name.
-