Permalink
Fetching contributors…
Cannot retrieve contributors at this time
553 lines (464 sloc) 25.9 KB

leofs-0.14.9

Features and Improvements for LeoFS

  • Improved
    • Supported Erlang/OTP R16B02
    • Increased performance of leo_redundant_manager

Used Libraries

leofs-0.14.8

Bugs Fixed

  • Gateway
    • Deletion failed with s3cmd in the case of not existed a targate file
    • OOM with access-log output
  • Manager
    • Start failed in the case of not reached the message from the Manager to storage-node(s)
    • Not removed a member(storage-node) when its status was 'attached'

Used Libraries

leofs-0.14.7

Features and Improvements for LeoFS

  • Improved
    • Gateway
      • Able to output access-log

Used Libraries

leofs-0.14.6

Features and Improvements for LeoFS

  • Improved

    • Able to set the max number of requests allowed in a single keep-alive session
  • Bugs Fixed

    • Gateway
      • Fixed to respond HTTP-headers properly when using disk-cache
    • Manager
      • Fixed changed status from running to restarted when restarted leo-manager

Used Libraries

leofs-0.14.5

Features and Improvements for LeoFS

  • Bugs Fixed
    • Gateway
      • Gateways may respond an HTTP response with a wrong MIME type as "plain/text" when requested an uppercase filename extension such as ".JPG" and ".PNG" from clients
      • Fixed an error handlings properly when processing a large file with disk cache

Used Libraries

leofs-0.14.4

Features and Improvements for LeoFS

  • Improved

Used Libraries

leofs-0.14.3

Features and Improvements for LeoFS

  • Improved
    • Improved redundant-manager's performance when retrieving redundancies by up to 20%
      • Changed method of retrieving redundancies from ets to worker-processes
    • Make a parameter of consumption of queue's message interval into the application config
    • Replace obsolete functions crypto:sha|md5 with crypto:hash

Bugs Fixed

  • Manager
    • Some problem with bucket names format
  • Storage
    • The number of queue messages can be wrong value
      • Needed to call bitcask:merge/1 after relaunch the storage process
    • Compaction-function can fail - "did not match data" when a target file is corrupted

Used Libraries

leofs-0.14.2

Features and Improvements for LeoFS

  • New Features
    • Supported rack-awareness replica placement
  • Improved
    • Gateway
      • Improved cache-libs, which are mem-cache and disc-cache.
        • Refactoring codes - Getting rid of duplicate codes to "libcutil"
        • Using file:sendfile/2 when retrieving an object from disk cache(dcerl)
      • Improve S3-API's compatibility
        • Supported "Range request API" for a large object
        • Supported retrieving object-list
      • Upgrade Cowboy to v0.8.5
    • Storage
      • Improved data-compaction
        • Performance tuning
        • Avoid polluting OS page caches during compaction
      • Modified the launch process
        • Able to retry connection with manager(s)
    • Manager
      • Supported to Able to remove gateway-node from manager's console when the state of specified node is ‘stop’
      • Supported system-data backup/restore (mnesia)
      • Supported taking over manager(s) for affected hardware failure
      • Supported taking over storage-node when having attach/detach nodes in operation
        • Able to take over from detach-node's files to attach-node

Bugs Fixed

  • Gateway
    • Reply empty response unintentionally when using some s3-clients

Used Libraries

leofs-0.14.1-1

Features and Improvements for LeoFS

  • Improved
    • Gateway
      • Upgrade Ranch and Cowboy to latest version

Bugs Fixed

  • Gateway
    • Reply empty response unintentionally when using some s3 clients
      • "leo_gateway" stored a "0-byte length object" into the cache when put an large-object(over 5MB) from clients. So We supported to remove a registered object from the cache.

leofs-0.14.1

Features and Improvements for LeoFS

  • Improved
    • Gateway
      • Commonize request-handler for easily creating APIs
      • Improve cache-controller which was replaced from ecache to leo_cache
      • Improve performance
        • Upgrade Ranch and Cowboy to latest version
    • Storage
      • Improved phased data-compaction
        • To realize non-blocking data-compaction
          • before: Making compaction per an "object-storage-file"
          • after: Making compaction per some an object
    • Manager/Storage
      • Implement judgment of precondition in rebalance-comamnd
      • Implemented recover-command in Manager
        • To realize:
          • synchronize a file: recover file ${file-path}
          • recover a target-node files: recover node ${storage-node}
          • recover target-node RING: recover ring ${storage-node}

Bugs Fixed

  • NOT worked s3cmd by degraded
  • Stored wrong file-path with REST-API by degraded

Used Libraries

leofs-0.14.0

Features and Improvements for LeoFS

  • Improved
    • Upgrade Cowboy from 0.6.2(foked/add patches) to 0.8.2(original)
      • Improved put/get large-object performance
    • Support recovery of manager-status-monitor after network partition occured

Bugs Fixed

  • Recovery objects (Storage)
    • Ignore detach-status from replication-message
  • Not handle compaction-start command invalid argument on Manager-console

Used Libraries

leofs-0.14.0-RC2

Features and Improvements for LeoFS

  • Improved
    • Improved cache-mechanism (Gateway)
      • Support Layered Cache, First cache-layer is RAM and Secondary cache-layer is SSD (or HDD)
      • This version realized that cache-destination is decided by object-size

Bugs Fixed

  • Incorrect judgement of active storage-node
    • Possibility of including node(s) of attached (status)

Used Libraries

leofs-0.14.0-RC1

Features and Improvements for LeoFS

  • Improved
    • Improved cache-mechanism (Gateway)
      • According to improving efficiency of object-cache-pool, LeoFS-Gateway was able to store an object up to 8MB into the RAM because Gateway can cache chunked objects
    • Improved data-compaction related processes (Storage, Manager)
      • Realized compact a part of data-storage, which called "phased data-compaction"
      • Realized suspend and resume data-compaction, also able to comfirm status of processes
    • Supported recovery from temporally network unlink #1 (Storage, Manager)
      • Automatically recover target-node(s), which status transition from stop to running

Bugs Fixed

  • Cannot consume queueing-message(s) when storage cluster has a detached node
    • Resolved that messages of a detached node ignored

Used Libraries