org.apache.sling.distribution.journal.impl.publisher

Class PackageRepo

java.lang.Object

org.apache.sling.distribution.journal.impl.publisher.PackageRepo

@ParametersAreNonnullByDefault
public class PackageRepo
extends Object

Manages the binary content of DistributionPackages. If they are too big to fit in a journal message then they are written to the blob store. It also offers cleanup functionality to remove the data when it is not needed anymore.

Constructor Summary

Constructors

Constructor and Description
PackageRepo()

Method Summary

Modifier and Type	Method and Description
void	cleanup() The cleanup algorithm is based on the head and tail offsets fetched from the package topic.
String	store(org.apache.sling.api.resource.ResourceResolver resolver, org.apache.sling.distribution.packaging.DistributionPackage disPkg)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

PackageRepo

public PackageRepo()

Method Detail

store

@Nonnull
public String store(org.apache.sling.api.resource.ResourceResolver resolver,
                             org.apache.sling.distribution.packaging.DistributionPackage disPkg)
                      throws org.apache.sling.distribution.common.DistributionException

Throws:

org.apache.sling.distribution.common.DistributionException

cleanup

public void cleanup()

The cleanup algorithm is based on the head and tail offsets fetched from the package topic. The implementation is robust and fits any topic retention policy setting, without requiring any explicit configuration matching the actual retention policy. It is also robust against potential instance clock misconfiguration/synchronisation issues. This comes at the expense of saving offsets in the repository. It should be an acceptable cost, given that the vast majority of packages are not saved in the repository on the first place and thus don't need cleanup. The cleanup does this: * The packages with an offset smaller than the head offset can be removed because they are no longer referenced by the package topic. * The packages with no offset are set the tail offset. This relies on the session being shielded from new packages that might be created in parallel.