v2.3.2

• support admin interface create_app #179
• support new batch_get for different hashKey #175
• fix thread-deadlock #168

v2.3.0

• support partition split #165
• some function refactor&fix #166 #164 #164
• update to git action #163

v2.2.0

• support security certification #138 #137 #136 #139 #135 #134 #132 #145
• enhance and update api
  • support compress and uncompress in table options #126
  • use Batch utils to replace batch api #129
  • make sure return "maxFetchCount" record #156
  • fix scan next can't recover bug #156 #148
• other funtions and code optimization #117 #153 #155

v2.0.0

Summary

Backup-request

• feat: implement backup request (#93)
• refactor(rpc): refactor request meta to adapt to new version of rpc protocol (#82)

Requirement: Pegasus server >= 2.0.0

Backup-request allows users to read from the secondary replica when latency from primary exceeds a preconfigured threshold.
It could help reduce the tail latency (p999 or higher) and achieve more stable performance.

Related discussion:

• apache/incubator-pegasus#251
• #84

Docs: http://pegasus.apache.org/administration/backup-request

Scan-limit

• feat: handle scan iteration exceed limit and refactor code (#96)

Requirement: Pegasus server >= 1.12.3

Other new features

• feat: forbid large-size-value written to pegasus server (#95)
• feat: add configuration for meta query timeout (#103)
• improvement: add detailed information when operation has exception (#106)

Bugfix

• Release streams resource while loading config (#113)
• fix: startSortKey can't pass null (#108)
• fix: use origin expireNanoTime to init new clientRequestRound (#105)
• fix: make SetItem Serializable (#101)
  This fix resolves the problem that writing to Pegasus in Spark will raise an exception since SetItem is unserializable.

Misc

• refactor: optimize perf-counter (#98)
• refactor: delete unused log in TableHandler (#87)
• refactor: reduce parameters in construction of internal Table and Cluster (#104)
• build: add maven-shade for thrift (#71)

Compatibility Note

v2.0.0 is only for >= Pegasus 2.0, otherwise, it will have compatibility problem which throw ERR_SESSION_RESET exception

1.11.10-thrift-0.11.0-inlined-release

Summary

• Add info for ERR_SESSION_RESET when querying meta failed (#66)
• Check the value of ttl (#69)
• Add maven-shade for zkclient&guava (#72)
• Use operation_timeout as socket connect timeout (#73)
• Bound minimal value for socket connect timeout (#74)
• Add check of invalid ttl value for checkAndSet interface (#75)
• Handle potential NullPointerException on RequestEntry.timeoutTask (#78)
• Close session when unable to connect (#81)
• Prevent session to be actively closed when it gets response from server (#86)

1.11.9-thrift-0.11.0-inlined-release

Summary

• Support client warmup to prepare connections with replica-server (#26)(#38)
• Add more exception info on synchronous APIs (#62)
• Add package version prefixed in PException (#63)
• Fix block bug of InetAddressValidator.getInstance() (#64)

1.11.8-thrift-0.11.0-inlined-release

Summary

• Add delRange api (#48)
• Add more exception information when the client's RPC request fails(#59)
• Support resolving meta host through DNS (#60)

1.11.7-thrift-0.11.0-inlined-release

Summary

Stability Fixes

• Upgrade netty version from 4.0.33.Final to 4.0.56.Final (#50)
• Ensure if response decode failed, client can be informed (#37)
• Bugfix on zookeeper client, because ZkClient was not closed after used. (#46)
• Fix batchSet implementation for the potential risk of dead-wait. (#51, #52)
• Set error to ERR_SESSION_RESET to trigger meta query while the session is not responding. This fix resolved the problem client can't recover from replica-server failure. (#54)

New features

• Support createClient API with ClientOptions, for users who don't use local file or Zookeeper to store configuration. (#49)
• Adapt to partition split. (#44)

Continuous Integration

• Support Travis-CI. (#47)
• Support code-format using checkstyle. (#35, #36, #40, #43)

1.11.5-thrift-0.11.0-inlined-release

• 支持客户端向不在配置中的 Meta Server 请求路由表：#28，#27

• 客户端重复超时多次就断开连接：#32

1.11.4-thrift-0.11.0-inlined-release

• 在PegasusCli工具中增加copy_data命令，用于复制表数据，并支持在复制流中对数据进行压缩/解压缩
• 将rpc header中client_timeout设置为正确的值，让server端可以获得客户端的超时时间