Key HBase Configuration Parameters and Production Recommendations (HBase 1.1.2)

Region : hbase.hregion.max.filesize defaults to 10 GB; in production it is recommended to set it to 50‑80 GB to balance compaction overhead and region split frequency.

BlockCache : The article advises using LRUBlockCache for < 20 GB memory and BucketCache (off‑heap) for larger caches. file.block.cache.size (default 0.4) controls LRU cache size as a fraction of JVM heap, while hbase.bucketcache.size defines off‑heap capacity; recommended values are 0.05‑0.1 for LRU and appropriate off‑heap size based on physical memory.

Memstore : hbase.hregion.memstore.flush.size defaults to 128 MB; increase to 256 MB if flushes are frequent and memory is ample. hbase.hregion.memstore.block.multiplier default 4 (often set to 5). hbase.regionserver.global.memstore.size defaults to 0.4; set to 0.6‑0.65 for off‑heap mode. hbase.regionserver.global.memstore.lowerLimit stays at 0.95. hbase.regionserver.optionalcacheflushinterval defaults to 1 hour; production often uses 10 hours to reduce small‑file generation.

Compaction : Key parameters include hbase.hstore.compactionThreshold (default 3, raise to 5‑10 for high write QPS), hbase.hstore.compaction.max (default 10, usually 2‑3× threshold), thread throttling settings, and hbase.hstore.blockingStoreFiles (default 10, raise to 100 to avoid write blocking). Major compaction interval hbase.hregion.majorcompaction defaults to one week; for large tables it is often disabled (set to 0).

HLog : hbase.regionserver.maxlogs defaults to 32, often increased per HBASE‑14951. hbase.regionserver.hlog.splitlog.writer.threads defaults to 3; production clusters typically raise to 10 to speed recovery.

Call Queue : hbase.regionserver.handler.count defaults to 30; production values range 100‑200. hbase.ipc.server.callqueue.handler.factor , hbase.ipc.server.callqueue.read.ratio , and hbase.ipc.server.call.queue.scan.ratio default to 0 and should be tuned based on read/write workload distribution.

Other Important Settings : Enable online schema updates with hbase.online.schema.update.enable (set true), turn on quota ( hbase.quota.enabled ) and snapshots ( hbase.snapshot.enabled ) for production, adjust zookeeper.session.timeout (default 180 s) in line with ZK server settings, enable multi‑update via hbase.zookeeper.useMulti , and configure coprocessor classes for security and token handling.

The article emphasizes that most HBase parameters remain at defaults; only a focused subset needs adjustment based on hardware, network, and workload characteristics, helping newcomers quickly grasp the rationale behind each setting.