Database Reference
In-Depth Information
Region server metrics
In an HBase cluster, region servers are very crucial components that participate in
the data read and write path. Hence, it exposes a large number of metrics for each
part of the HBase internal architecture. The following list of metrics is provided
for detailed information:
•
Block cache metrics
: A block cache stores the loaded storage blocks from
the low-level HFiles and are kept in memory until it is full. It exposes the
following metrics:
° Count metric: This is the number of blocks currently in the cache
° Size metric: This is the occupied Java heap space
° Free metric: This is the remaining heap for the cache
° Evicted metric: This presents the counts of blocks that had to be
removed from the block cache because of heap size constraints
with JVM
•
Compaction metrics
: A region server performs the asynchronous
housekeeping task of compacting the storage iles. During this process,
it exposes the following metrics:
° Compaction size metric and compaction time metric: This provides
the total size (in bytes) of the storage files that have been compacted
and the time taken by the compaction operation
° Compaction queue size metric: This provides the count of files
queued up at the region server and waiting for the compaction
operation to be triggered
•
MemStore metrics
: Data contained within MemStore on the region server is
written to the disk via lush operations. The HBase region server exposes the
following metrics for MemStore:
° Memstore size metric: This provides the total heap space occupied
by all MemStores for the server across all the available regions
in megabytes
° Flush queue size metric: This provides the number of regions
ready for being flushed
° Flush size metric: This provides the total size of the MemStore
being flushed
° Flush time metric: This provides the total time taken for flushing
the MemStore
