com.gemstone.gemfire.cache.partition

Class PartitionRegionHelper

java.lang.Object

com.gemstone.gemfire.cache.partition.PartitionRegionHelper

public final class PartitionRegionHelper
extends Object

Utility methods for handling partitioned Regions, for example during execution of Functions on a Partitioned Region.

Example of a Function using utility methods:

  public Serializable execute(FunctionContext context) {
     if (context instanceof RegionFunctionContext) {
         RegionFunctionContext rc = (RegionFunctionContext) context;
         if (PartitionRegionHelper.isPartitionedRegion(rc.getDataSet())) {
             Region efficientReader =
              PartitionRegionHelper.getLocalDataForContext(rc);
             efficientReader.get("someKey");
             // ...
         }
      }
  // ...
 

Since:

6.0

See Also:

FunctionService.onRegion(Region)

Method Summary

Methods

Modifier and Type	Method and Description
static void	assignBucketsToPartitions(Region<?,?> region) Decide which partitions will host which buckets.
static <K,V> Set<DistributedMember>	getAllMembersForKey(Region<K,V> r, K key) Get all potential owners for a key.
static Map<String,Region<?,?>>	getColocatedRegions(Region<?,?> r) Given a partitioned Region, return a map of colocated Regions.
static Map<String,Region<?,?>>	getLocalColocatedRegions(RegionFunctionContext c) Given a RegionFunctionContext for a partitioned Region, return a map of colocated Regions with read access limited to the context of the function.
static <K,V> Region<K,V>	getLocalData(Region<K,V> r) Given a partitioned Region return a Region providing read access limited to the local heap, writes using this Region have no constraints and behave the same as a partitioned Region.
static <K,V> Region<K,V>	getLocalDataForContext(RegionFunctionContext c) Given a RegionFunctionContext for a partitioned Region, return a Region providing read access limited to the function context.
static <K,V> Region<K,V>	getLocalPrimaryData(Region<K,V> r) Given a partitioned Region return a Region providing read access to primary copy of the data which is limited to the local heap, writes using this Region have no constraints and behave the same as a partitioned Region.
static Set<PartitionRegionInfo>	getPartitionRegionInfo(Cache cache) Gathers a set of details about all partitioned regions in the local Cache.
static PartitionRegionInfo	getPartitionRegionInfo(Region<?,?> region) Gathers details about the specified partitioned region.
static <K,V> DistributedMember	getPrimaryMemberForKey(Region<K,V> r, K key) Get the current primary owner for a key.
static <K,V> Set<DistributedMember>	getRedundantMembersForKey(Region<K,V> r, K key) Get all potential redundant owners for a key.
static boolean	isPartitionedRegion(Region<?,?> r) Test a Region to see if it is a partitioned Region
static <K> void	moveBucketByKey(Region<K,?> region, DistributedMember source, DistributedMember destination, K key) Moves the bucket which contains the given key from the source member to the destination member.
static RebalanceResults	moveData(Region<?,?> region, DistributedMember source, DistributedMember destination, float percentage) Moves data from the source member to the destination member, up to the given percentage of data (measured in bytes).

Methods inherited from class java.lang.Object

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

Method Detail

getColocatedRegions

public static Map<String,Region<?,?>> getColocatedRegions(Region<?,?> r)

Given a partitioned Region, return a map of colocated Regions. Given a local data reference to a partitioned region, return a map of local colocated Regions. If there are no colocated regions, return an empty map.

Parameters:

r - a partitioned Region

Returns:

an unmodifiable map of region name to Region

Throws:

IllegalStateException - if the Region is not a partitioned Region

Since:

6.0

isPartitionedRegion

public static boolean isPartitionedRegion(Region<?,?> r)

Test a Region to see if it is a partitioned Region

Parameters:

r -

Returns:

true if it is a partitioned Region

Since:

6.0

getPartitionRegionInfo

public static Set<PartitionRegionInfo> getPartitionRegionInfo(Cache cache)

Gathers a set of details about all partitioned regions in the local Cache. If there are no partitioned regions then an empty set will be returned.

Parameters:

cache - the cache which has the regions

Returns:

set of details about all locally defined partitioned regions

Since:

6.0

getPartitionRegionInfo

public static PartitionRegionInfo getPartitionRegionInfo(Region<?,?> region)

Gathers details about the specified partitioned region. Returns null if the partitioned region is not locally defined.

Parameters:

region - the region to get info about

Returns:

details about the specified partitioned region

Since:

6.0

assignBucketsToPartitions

public static void assignBucketsToPartitions(Region<?,?> region)

Decide which partitions will host which buckets. Gemfire normally assigns buckets to partitions as needed when data is added to a partitioned region. This method provides way to assign all of the buckets without putting any data in partition region. This method should not be called until all of the partitions are running because it will divide the buckets between the running partitions. If the buckets are already assigned this method will have no effect. This method will block until all buckets are assigned.

Parameters:

region - The region which should have it's buckets assigned.

Throws:

IllegalStateException - if the provided region is something other than a partitioned Region

Since:

6.0

getPrimaryMemberForKey

public static <K,V> DistributedMember getPrimaryMemberForKey(Region<K,V> r,
                                             K key)

Get the current primary owner for a key. Upon return there is no guarantee that primary owner remains the primary owner, or that the member is still alive.

This method is not a substitute for Region.containsKey(Object).

Parameters:

r - a PartitionedRegion

key - the key to evaluate

Returns:

the primary member for the key, possibly null if a primary is not yet determined

Throws:

IllegalStateException - if the provided region is something other than a partitioned Region

Since:

6.0

getRedundantMembersForKey

public static <K,V> Set<DistributedMember> getRedundantMembersForKey(Region<K,V> r,
                                                     K key)

Get all potential redundant owners for a key. If the key exists in the Region, upon return there is no guarantee that key has not been moved or that the members are still alive.

This method is not a substitute for Region.containsKey(Object).

This method is equivalent to: DistributedMember primary = getPrimaryMemberForKey(r, key); Set allMembers = getAllMembersForKey(r, key); allMembers.remove(primary);

Parameters:

r - a PartitionedRegion

key - the key to evaluate

Returns:

an unmodifiable set of members minus the primary

Throws:

IllegalStateException - if the provided region is something other than a partitioned Region

Since:

6.0

getAllMembersForKey

public static <K,V> Set<DistributedMember> getAllMembersForKey(Region<K,V> r,
                                               K key)

Get all potential owners for a key. If the key exists in the Region, upon return there is no guarantee that it has not moved nor does it guarantee all members are still alive.

This method is not a substitute for Region.containsKey(Object).

Parameters:

r - PartitionedRegion

key - the key to evaluate

Returns:

an unmodifiable set of all members

Throws:

IllegalStateException - if the provided region is something other than a partitioned Region

Since:

6.0

getLocalColocatedRegions

public static Map<String,Region<?,?>> getLocalColocatedRegions(RegionFunctionContext c)

Given a RegionFunctionContext for a partitioned Region, return a map of colocated Regions with read access limited to the context of the function.

Writes using these Region have no constraints and behave the same as a partitioned Region.

If there are no colocated regions, return an empty map.

Parameters:

c - the region function context

Returns:

an unmodifiable map of region name to Region

Throws:

IllegalStateException - if the Region is not a partitioned Region

Since:

6.0

getLocalDataForContext

public static <K,V> Region<K,V> getLocalDataForContext(RegionFunctionContext c)

Given a RegionFunctionContext for a partitioned Region, return a Region providing read access limited to the function context.
Returned Region provides only one copy of the data although redundantCopies configured is more than 0. If the invoking Function is configured to have optimizeForWrite as true,the returned Region will only contain primary copy of the data.

Writes using this Region have no constraints and behave the same as a partitioned Region.

Parameters:

c - a functions context

Returns:

a Region for efficient reads

Throws:

IllegalStateException - if RegionFunctionContext.getDataSet() returns something other than a partitioned Region

Since:

6.0

getLocalData

public static <K,V> Region<K,V> getLocalData(Region<K,V> r)

Given a partitioned Region return a Region providing read access limited to the local heap, writes using this Region have no constraints and behave the same as a partitioned Region.

Parameters:

r - a partitioned region

Returns:

a Region for efficient reads

Throws:

IllegalStateException - if the provided region is something other than a partitioned Region

Since:

6.0

getLocalPrimaryData

public static <K,V> Region<K,V> getLocalPrimaryData(Region<K,V> r)

Given a partitioned Region return a Region providing read access to primary copy of the data which is limited to the local heap, writes using this Region have no constraints and behave the same as a partitioned Region.

Parameters:

r - a partitioned region

Returns:

a Region for efficient reads

Throws:

IllegalStateException - if the provided region is something other than a partitioned Region

Since:

6.5

moveBucketByKey

public static <K> void moveBucketByKey(Region<K,?> region,
                       DistributedMember source,
                       DistributedMember destination,
                       K key)

Moves the bucket which contains the given key from the source member to the destination member. The bucket will be fully transferred once this method is complete, if the method does not throw an exception.

All keys which exist in the same bucket will also be moved to the new node.

Any data in colocated regions that are colocated with this key will also be moved.

This method allows direct control of what data to move. To automatically balance buckets, see ResourceManager.createRebalanceFactory()

Parameters:

region - The region in which to move the bucket. Data in regions colocated with this region will also be moved.

source - A member that is currently hosting this bucket. The bucket is moved off of this member.

destination - A member that is not currently hosting this bucket, but has the partitioned region defined. The bucket is moved to this member.

key - A key which maps to the bucket to move. This key does not actually need to exist in the region, but if using a PartitionResolver the resolver should be able to get the routing object from this key to determine the bucket to move.

Throws:

IllegalStateException - if the bucket is not present on the source, if the source or destination are not valid members of the system, if the destination already hosts a copy of the bucket, or if the bucket does not exist.

Since:

7.1

moveData

public static RebalanceResults moveData(Region<?,?> region,
                        DistributedMember source,
                        DistributedMember destination,
                        float percentage)

Moves data from the source member to the destination member, up to the given percentage of data (measured in bytes). The data will be fully transferred once this method is complete, if the method does not throw an exception. The percentage is a percentage of the amount of data in bytes on the source member for this region.

If this region has colocated regions, the colocated data will also be moved. The total amount of data in all colocated regions will be taken into consideration when determining what percentage of data will be moved.

It may not be possible to move data to the destination member, if the destination member has no available space, no bucket smaller than the given percentage exists, or if moving data would violate redundancy constraints. If data cannot be moved, this method will return a RebalanceResult object with 0 total bucket transfers.

This method allows direct control of what data to move. To automatically balance buckets, see ResourceManager.createRebalanceFactory()

Parameters:

region - The region in which to move data. Data in regions colocated with this region will also be moved.

source - A member that is currently hosting data. The bucket is moved off of this member.

destination - A member that that has the partitioned region defined. Data is moved to this member.

percentage - the maximum amount of data to move, as a percentage from 0 to 100.

Returns:

A RebalanceResult object that contains information about what what data was actually moved.

Throws:

IllegalStateException - if the source or destination are not valid members of the system.

IllegalArgumentException - if the percentage is not between 0 to 100.

Since:

7.1