Skip to content

2.8.2.1.rc2

Pre-release
Pre-release
Compare
Choose a tag to compare
@berthoug berthoug released this 16 Nov 10:00
· 2 commits to 2.8.2.1 since this release
2.8.2.1.rc2
f7754aa 
    Release notes - hops - Version 2.8.2.1

Bug

  • [HOPS-14] - remove thread in NamenodeSelector
  • [HOPS-16] - Transactions deadlock exception is thrown when hundreds of clients simultaneously read the same file.
  • [HOPS-20] - Lease manager dies on Transient exception.
  • [HOPS-21] - Poor Synchronous implementation of Namenode Selector in HopsFS 2.7.3 causes 80-90% drop in throughput
  • [HOPS-45] - Slow LS operation on directories with large number of immediate descendants
  • [HOPS-47] - TestUsersGroups fail due to wrong order of list elements
  • [HOPS-48] - FATAL UniqueKeyConstraintViolationException During Namenode Creation
  • [HOPS-51] - hdfs native_tests failing
  • [HOPS-56] - setOwner fails with ForeignKeyConstraintViolationException due to outdated cache
  • [HOPS-59] - HopsSSLSocketFactory assumes there is a single proxy superuser
  • [HOPS-67] - Incoming data are not processed in Server
  • [HOPS-69] - Delete directoy does not work non leader namenode when quota is enabled
  • [HOPS-71] - Fix permission issue when writing to Cgroup device subsystem
  • [HOPS-75] - Fix Log Aggregation in Yarn
  • [HOPS-76] - ProcfsBasedProcessTree cannot parse PID process tree info
  • [HOPS-81] - Missing version specifier in hadoop-common dependency causes warnings
  • [HOPS-82] - Hops-GPU should handle cases when a subset of GPUs are not working correctly
  • [HOPS-84] - hadoop-common pom.xml must include hops-gpu-management as dependency
  • [HOPS-85] - Change collate for all the remaining yarn_* tables
  • [HOPS-89] - Owner not correctly set in Cgroup hierarchy
  • [HOPS-90] - Cgroups broken due to bug in container-executor
  • [HOPS-91] - DFSClient update NN list as superuser
  • [HOPS-92] - Symbol lookup error crashes NodeManager
  • [HOPS-93] - Refactor hops-gpu-manage and hops-gpu-management-impl-nvidia
  • [HOPS-103] - Concurrency bug in blockreporting causes iterator infinite loop
  • [HOPS-107] - Small file's read fails when a decommissioned node is selected as a phanton datanode
  • [HOPS-108] - Get HDFS username from environment variable in libhdfs
  • [HOPS-110] - Change column type of small files tables from varchar to varbinary
  • [HOPS-111] - Support set replication api for small files
  • [HOPS-112] - Implement seek api for small files
  • [HOPS-114] - Can't build without -Pndb since [HOPS-103]
  • [HOPS-116] - Spark applications fail when small fiels are read using read(ByteBuffer) api
  • [HOPS-123] - Guard for no data in phantom blocks
  • [HOPS-127] - pom <repositories> and <distributionManagement> badly configured
  • [HOPS-128] - libhopsyarn pointing to wrong table
  • [HOPS-130] - Container log remaining in DB for containers stoped just after creation

Sub-task

  • [HOPS-41] - NPE in Rename for metaEnabled dirs and subdirs
  • [HOPS-42] - Optimize logging of rename in the same meta enabled directory
  • [HOPS-43] - Delete after rename of a metaEnabled directory fail
  • [HOPS-55] - Refactoring block report processing
  • [HOPS-58] - Correcting code quality issues in FSNamesystem
  • [HOPS-60] - Remove remains of Federation from HopsFS
  • [HOPS-62] - Replace Block Report RPC with Bucketized RPC
  • [HOPS-63] - Create a HashBucket entity
  • [HOPS-64] - Replace the block reporting mechanism with the hash-based one

New Feature

  • [HOPS-2] - Improve Small File Performance in HopsFS
  • [HOPS-4] - Uber style resource price
  • [HOPS-7] - TLS support for Hadoop RPC/IPC
  • [HOPS-11] - GPU support for HopsYARN (Umbrella)
  • [HOPS-12] - Create an abstraction layer for GPU devices
  • [HOPS-13] - Create an implementation layer for GPU devices
  • [HOPS-19] - Truncate support in HDFS
  • [HOPS-23] - avoid running app to be able to access other app data
  • [HOPS-124] - libhopsyarn and libhopsnvml shouldn't be included in the pom files.

Improvement

  • [HOPS-17] - Optimizing addBlock and incrementalBlockReport RPC for large files
  • [HOPS-22] - StorageCallPrevented Exception must not be swallowed
  • [HOPS-26] - Clean up old MetadataLog code in TransactionalRequestHandler
  • [HOPS-29] - ByteBuffer Mem Leak Problem
  • [HOPS-30] - Expose Parameters For TX Retry
  • [HOPS-31] - Add support to GPUs for hops-metadata-dal
  • [HOPS-32] - Add support to GPUs for hops-metadata-dal-impl-ndb
  • [HOPS-34] - HdfsFileStatus should return the fileId
  • [HOPS-35] - Optimizing logging of MetaEnabled directories
  • [HOPS-37] - Append To file using HDFS clients (version >= v2.4) fails due to NPE on the HDFS clinet side
  • [HOPS-49] - Making HopsFS and the DAL Java 8 syntax compliant
  • [HOPS-54] - Remove unnecessary debug printouts from dal-impl-ndb
  • [HOPS-57] - Hash based block-reporting for improved block reporting speed
  • [HOPS-61] - Block Management Refactoring
  • [HOPS-65] - update to follow hadoop 2.8.2 updates
  • [HOPS-66] - Add a script to upgrade Clusterj with hopsfs fixes and Clusterj Native in Kompics repo
  • [HOPS-68] - Pass the keystore and truststore password along with the material
  • [HOPS-79] - Apply patch YARN-7083 from upstream fixing Log Aggregation
  • [HOPS-83] - Cgroups hierarchy should be created automatically
Assets 2
Loading