Package org.apache.hadoop.hbase.client

Class AsyncTableRegionLocatorImpl

java.lang.Object
org.apache.hadoop.hbase.client.AsyncTableRegionLocatorImpl
All Implemented Interfaces:
AsyncTableRegionLocator
@Private class AsyncTableRegionLocatorImpl extends Object implements AsyncTableRegionLocator
The implementation of AsyncRegionLocator.

  • Field Details

  • Constructor Details

  • Method Details

    • getName

      public TableName getName()
      Description copied from interface: AsyncTableRegionLocator
      Gets the fully qualified table name instance of the table whose region we want to locate.
      Specified by:
      getName in interface AsyncTableRegionLocator

    • getRegionLocation

      public CompletableFuture<HRegionLocation> getRegionLocation(byte[] row, int replicaId, boolean reload)
      Description copied from interface: AsyncTableRegionLocator
      Finds the region with the given replicaId on which the given row is being served.

      Returns the location of the region with the given replicaId to which the row belongs.

      Specified by:
      getRegionLocation in interface AsyncTableRegionLocator
      Parameters:
      row - Row to find.
      replicaId - the replica id of the region
      reload - true to reload information or false to use cached information

    • getAllRegionLocations

      public CompletableFuture<List<HRegionLocation>> getAllRegionLocations()
      Description copied from interface: AsyncTableRegionLocator
      Retrieves all of the regions associated with this table.

      Usually we will go to meta table directly in this method so there is no reload parameter.

      Notice that the location for region replicas other than the default replica are also returned.

      Specified by:
      getAllRegionLocations in interface AsyncTableRegionLocator
      Returns:
      a List of all regions associated with this table.

    • getRegionLocations

      public CompletableFuture<List<HRegionLocation>> getRegionLocations(byte[] row, boolean reload)
      Description copied from interface: AsyncTableRegionLocator
      Find all the replicas for the region on which the given row is being served.
      Specified by:
      getRegionLocations in interface AsyncTableRegionLocator
      Parameters:
      row - Row to find.
      reload - true to reload information or false to use cached information
      Returns:
      Locations for all the replicas of the row.

    • getRegionLocationsPage

      public CompletableFuture<List<HRegionLocation>> getRegionLocationsPage(byte[] startKey, int limit)
      Description copied from interface: AsyncTableRegionLocator
      Bulk lookup of region locations from hbase:meta in a single RPC, starting at startKey (region start-key boundary, inclusive) and returning at most limit regions in start-key order.

      The returned list includes all replicas of each region (matching AsyncTableRegionLocator.getAllRegionLocations()), and the result is also written to the connection's region location cache.

      Ordering: regions are returned in ascending region start-key order (the natural order of hbase:meta rows for a single table). Within each region, replicas are returned in ascending replica-id order (replica 0, then 1, then 2, ...). Split parents are filtered out, which may cause a page to contain fewer than limit regions but never disturbs ordering of the survivors.

      To page through all regions of a table, call repeatedly passing last.getRegion().getEndKey() as the next startKey, where last is the final element of the previous response. All replicas of a region share the same RegionInfo, so the last entry's end key is the correct cursor regardless of which replica it is. Pass null for the first call. Stop paging when the returned list is empty or when the last region's end key is HConstants.EMPTY_END_ROW (zero-length) - that signals the end of the table; passing it back in would re-scan from the beginning since by convention an empty start key means "from the first region".

      Unlike AsyncTableRegionLocator.getAllRegionLocations(), this method performs at most one RPC against hbase:meta per invocation, so its latency is bounded by limit rather than table size. The single-RPC behavior is best-effort: if the response would exceed hbase.client.scanner.max.result.size (default 2 MB), the server may split the slice across multiple ScannerNext RPCs. For typical meta row sizes and default caching this rarely fires, but callers passing large limit values against clusters with replicas or heavy meta rows should treat single-RPC as a soft guarantee, not absolute. Note that this method does not coordinate with other in-flight meta lookups on the connection - aggregate pacing across concurrent callers is the caller's responsibility.

      This method is optional. Implementations that cannot support paginated lookups will return a future that completes exceptionally with UnsupportedOperationException (the default behavior); callers should fall back to AsyncTableRegionLocator.getAllRegionLocations() in that case.

      Specified by:
      getRegionLocationsPage in interface AsyncTableRegionLocator
      Parameters:
      startKey - region start-key to begin scanning from (inclusive); null or empty starts from the first region
      limit - maximum number of regions to return. If <= 0, falls back to hbase.meta.scanner.caching - this is a SOFT cap on a single page, NOT "all regions"; tables larger than the cap still require the caller to keep paging via last.getRegion().getEndKey().
      Returns:
      up to limit HRegionLocations in start-key order, possibly empty when no more regions exist; errors are reported via the returned future

    • clearRegionLocationCache

      public void clearRegionLocationCache()
      Description copied from interface: AsyncTableRegionLocator
      Clear all the entries in the region location cache.

      This may cause performance issue so use it with caution.

      Specified by:
      clearRegionLocationCache in interface AsyncTableRegionLocator