v0.3.0

^{Release notes were automatically generated by Shipkit}

0.3.0

• 2019-08-22 - 2396 commits by 29 authors - published to
• Commits: Sivabalan Narayanan (592), Gopalakrishna Holla (484), Priyesh Narayanan (402), sriramsub (325), Casey Getz (99), Tianwei Li (96), Jay J Wylie (72), Ming Xia (71), Ze Mao (63), Yingyi Zhang (60), Terry Li (32), Rob Block (30), David Harju (29), Xun Yin (12), Sriram Subramanian (6), Ankur Agrawal (4), Dmitry Nikiforov (3), Jasper Siepkes (3), Justin Lin (3), André Laszlo (1), Bas van Schaik (1), Diego Mariani (1), Enrico Olivelli (1), heipacker (1), johnnyjian (1), Naresh Bafna (1), ohze.net (1), Shuai Yuan (1), Vincent Bernat (1)
• Fix directory deletion failure in DiskSpaceAllocator (#1244)
• Add config option to connect to Cosmos with DirectHttps mode (#1243)
• Use Timer instead of Histgram on some metrics (#1242)
• Enable shipkit release publishing (#1239)
• Fix gradle heap size configs (#1238)
• Misc fixes in cluster manager, replication and compaction manager (#1237)
• Clean up network related metrics. (#1236)
• Upgrade gradle, fix test logging and int test CPU usage (#1235)
• Implement helixAccountService with version history and blobs (#1234)
• Support dynamically removing store from storage manager (#1232)
• Fix broken builds by specifying Ubuntu trusty dist. (#1231)
• Run azure recovery locally (#1230)
• Implement get functionality for Azure Vcr (#1229)
• Introducing Ambry partition state model and update Helix tool (#1228)
• Create executor pool to take over network read/write tasks. (#1227)
• Support dynamically adding store to storage manager (#1226)
• Add cache of recently seen blobs in CloudBlobStore (#1225)
• Ensure blobId with prefix in request header is correctly handled (#1224)
• Set permissions for ambry data and operation files (#1223)
• Support QPS throttling in NettyPerfClient (#1220)
• Misc changes for adaptive operation tracker (#1218)
• Allow NettyPerfClient to connect multiple frontends (#1217)
• Implement VCR findEntriesSince method (#1216)
• Free PutOperation references earlier (#1215)
• Add more log infos in AmbryStateModel (#1214)
• Allow store to skip disk flush during shutdown if disk I/O error occurs (#1212)
• Cloud compactor schedule (#1211)
• Refactoring DiskSpaceAllocator to support dynamically add/remove store (#1210)
• Fix bug causing uploaded blobs to be empty (#1209)
• Downgrade metrics library (#1208)
• Fix partition lag metric typo (#1207)
• Shutdown request handler before network server in ambry server (#1206)
• Retry Cosmos requests after delay when they return 429 (#1205)
• Fix typo in partition lag metric name (#1204)
• Add off switch for replenishConnections and honor datanode health (#1203)
• Fixed some issues with Segment GETs (#1202)
• Fix potential cache incoherence issue. (#1201)
• Add Cosmos record even if blob was previously uploaded (#1200)
• Add configurable warm up for remote datacenter connections (#1198)
• Improve GetWritablePartitionIds Performance for PUT Requests (#1197)
• Improve OperationTracker to terminate when there are two NOT_FOUND from originating DC (#1196)
• Add per partition replication lag metrics. (#1195)
• Upgrade Helix library version and add auto maintenance mode in HelixVcrPopulateTool (#1194)
• Segmented get metrics (#1193)
• Azure container name must be lower case (#1192)
• Add support for querying and purging dead blobs in Azure (#1189)
• Add post blob size (#1188)
• Segmented blob GET (#1187)
• Maintain connection low watermark in NetworkClient (#1186)
• Use cluster name as prefix in Azure blob container (#1185)
• Introducing node-level/disk-level histogram in adaptive tracker (#1184)
• Use customized blobId string as Azure blob name. (#1181)
• Ensure all disconnected connections are handled by ResponseHandler (#1180)
• Add resource control enable/disable (#1179)
• Fix helix test issues. (#1177)
• Added HelixPopulateTool and tests. (#1176)
• Stitched blobs support different sized component blobs (#1175)
• Introducing partition-level histogram into adaptive tracker (#1174)
• Exclude timed out response from adaptive tracker's histogram (#1173)
• Add logs in CompositeClusterManager to debug writable partition mismatch (#1172)
• Introduce HelixVcrCluster to CloudBackupManager and refactor ReplicationEngine and ReplicaThread. (#1171)
• Minor fixes in IOException catch and metrics registry (#1170)
• Fix to correct #1165. (#1169)
• Fix NullPointerException and 'no more capacity' issues in blob store (#1168)
• Cloud tokens (#1166)
• Blobs get TTL upload should be uploaded to Cloud. (#1165)
• Make CompositeClusterManager more robust to NPEs and other errors (#1164)
• Avoid unnecessary round trips to Azure when uploading blobs (#1163)
• Fixing the intermittent Helix tool test failure (#1162)
• Introduce container metrics (#1161)
• Add more error message to help debug invalid offset (#1160)
• Refactor operation tracker ctor and introduce custom percentiles (#1159)
• Add HelixVcrCluster and HelixVcrClusterFactory. (#1158)
• Misc minor fixes in Helix tool, ReplicationEngine and Composite Manager (#1157)
• Create ZkServerWrapper to speed up test. (#1156)
• Set network proxy in config rather than jvm args (#1155)
• Combined cloud blob crypto service (#1154)
• Provide option to use GetBlobOperation for BlobInfo (#1153)
• Avoid calling clusterMap.getReplicas() in VCR m...

First version of Ambry

This is the first public release of Ambry. The following are supported -

1. PUT, GET and DELETE of objects
2. Encryption between data nodes
3. REST based frontend
4. Replication within and across DCs
5. Failure detection
6. Change capture at the storage and frontend