0.4.0 beta1 artifacts are now being voted on

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

0.4.0 beta1 artifacts are now being voted on

Jonathan Ellis-3
If you would like to review the artifacts while the Apache Incubator
PMC votes on them, you can find them at
http://people.apache.org/~eevans.  After the vote (assuming it is in
the affirmative) we will push them to the mirrors and website.

Changelog (from CHANGES.txt):

 * On-disk data format has changed to allow billions of keys/rows per
   node instead of only millions
 * Multi-keyspace support
 * Added option to fsync before acknowledging writes
 * Scan all sstables for all queries to avoid situations where
   different types of operation on the same ColumnFamily could
   disagree on what data was present
 * Configurable LRU cache for key lookups
 * Snapshot support via JMX
 * Thrift API has changed a _lot_:
    - removed time-sorted CFs; instead, user-defined comparators
      may be defined on the column names, which are now byte arrays.
      Default comparators are provided for UTF8, Bytes, Ascii, Long (i64),
      and UUID types.
    - removed colon-delimited strings in thrift api in favor of explicit
      structs such as ColumnPath, ColumnParent, etc.  Also normalized
      thrift struct and argument naming.
    - Added columnFamily argument to get_key_range.
    - Change signature of get_slice and get_slice_super to accept
      starting and ending columns as well as an offset.  (This allows use
      of indexes.)  Added "ascending" flag to allow reasonably-efficient
      reverse scans as well.  Removed get_slice_by_range as redundant.
    - Similarly, changed signature of get_slice_super.
    - get_key_range operates on one CF at a time
    - changed `block` boolean on insert methods to ConsistencyLevel enum,
      with options of NONE, ONE, QUORUM, and ALL.
    - added similar consistency_level parameter to read methods
    - column-name-set slice with no names given now returns zero columns
      instead of all of them.  ("all" can run your server out of memory.
      use a range-based slice with a high max column count instead.)
 * Removed the web interface. Node information can now be obtained by
   using the newly introduced nodeprobe utility.
 * More JMX stats
 * Remove magic values from internals (e.g. special key to indicate
   when to flush memtables)
 * Rename configuration "table" to "keyspace"
 * Moved to crash-only design; no more shutdown (just kill the process)
 * Lots of bug fixes

Issues remaining for 0.4 (we expect to address these by beta2):

All issues resolved during 0.4: