Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Newer
Older
100644 611 lines (499 sloc) 23.701 kb
a3e2926 @jbellis r/m binarymemtable
jbellis authored
1 1.0
2 ===
3
4 Upgrading
5 ---------
6 - the BinaryMemtable bulk-load interface has been removed. Use the
7 sstableloader tool instead.
64f9337 @jbellis r/m compaction_thread_priority setting
jbellis authored
8 - the compaction_thread_priority setting has been removed from
9 cassandra.yaml (use compaction_throughput_mb_per_sec to throttle
10 compaction instead)
a3e2926 @jbellis r/m binarymemtable
jbellis authored
11
5f49666 @jbellis update README and NEWS
jbellis authored
12 Features
13 --------
3e77792 @pcmanus Use only one parameters for compression options
pcmanus authored
14 - SSTable compression is supported through the 'compression_options'
15 parameter when creating/updating a column family. For instance, you can
16 create a column family Cf using compression (using the Snappy library)
17 in the CLI with:
18 create column family Cf with compression_options={sstable_compression: SnappyCompressor}
19 SSTable compression is not activated by default but can be activated or
20 deactivated at any time.
ec5c1a0 @jbellis update CHANGES, NEWS
jbellis authored
21 - Compressed SSTable blocks are checksummed to protect against bitrot
4dbda61 @jbellis update NEWS
jbellis authored
22 - New LevelDB-inspired compaction algorithm can be enabled by setting the
23 Columnfamily compaction_strategy=LeveledCompactionStrategy option.
24 Leveled compaction means you only need to keep a few MB of space free for
25 compaction instead of (in the worst case) 50%.
ec5c1a0 @jbellis update CHANGES, NEWS
jbellis authored
26 - Windows Service ("cassandra.bat install" to enable)
f6e7947 @jbellis update NEWS for #2034
jbellis authored
27 - Hinted Handoff has two major improvements:
28 - Hint replay is much more efficient thanks to a change in the data model
29 - Hints are created for all replicas that do not ack a write. (Formerly,
30 only replicas known to be down when the write started were hinted.)
31 This means that running with read repair completely off is much more
32 viable than before.
62cc775 @jbellis convenience workflow for replacing dead node
jbellis authored
33 - A dead node may be replaced in a single step by starting a new node
34 with -Dcassandra.replace_token=<token>. More details can be found at
35 http://wiki.apache.org/cassandra/Operations#Replacing_a_Dead_Node
3e77792 @pcmanus Use only one parameters for compression options
pcmanus authored
36 - It is now possible to repair only the first range returned by the
37 partitioner for a node with `nodetool repair -pr`. It makes it
38 easier/possible to repair a full cluster without any work duplication by
39 running this command on every node of the cluster.
ec5c1a0 @jbellis update CHANGES, NEWS
jbellis authored
40
41 Other
42 -----
43 - Hinted Handoff is substantially more robust, with the result that
44 when HH is enabled, repair only needs to be run if a node crashes.
45 - Because of this, read repair is disabled now by default on newly
46 created ColumnFamilies.
5f49666 @jbellis update README and NEWS
jbellis authored
47
a3e2926 @jbellis r/m binarymemtable
jbellis authored
48
652ab2a @pcmanus merge from 0.8
pcmanus authored
49 0.8.5
50 =====
51
52 Other
53 -----
54 - The sstableloader is now bundled with the debian package
55
56
0bbc22c @jbellis merge from 0.8
jbellis authored
57 0.8.4
58 =====
59
60 Upgrading
61 ---------
62 - Nothing specific to 0.8.4
63
64 Other
65 -----
66 - This release comes to fix a bug in counter that could lead to
67 (important) over-count.
68 - It also fixes a slight upgrade regression from 0.8.3. It is thus advised
69 to jump directly to 0.8.4 if upgrading from before 0.8.3.
70
71
d505bde @jbellis merge from 0.8
jbellis authored
72 0.8.3
f304c4c @pcmanus Updates for 0.7.8 release (changelog, news, version number)
pcmanus authored
73 =====
74
75 Upgrading
76 ---------
d505bde @jbellis merge from 0.8
jbellis authored
77 - Token removal has been revamped. Removing tokens in a mixed cluster with
78 0.8.3 will not work, so the entire cluster will need to be running 0.8.3
79 first, except for the dead node.
f304c4c @pcmanus Updates for 0.7.8 release (changelog, news, version number)
pcmanus authored
80
455dcc5 @jbellis merge from 0.8
jbellis authored
81 Features
82 --------
83 - It is now possible to use thrift asynchronous and
84 half-synchronous/half-asynchronous servers (see cassandra.yaml for more
85 details).
86 - It is now possible to access counter columns through Hadoop.
87
88 Other
89 -----
90 - This release fix a regression of 0.8 that can make commit log segment to
91 be deleted even though not all data it contains has been flushed.
92 Upgrades from 0.8.* is very much encouraged.
93
f304c4c @pcmanus Updates for 0.7.8 release (changelog, news, version number)
pcmanus authored
94
6652b5d @jbellis merge from 0.8
jbellis authored
95 0.8.2
96 =====
97
98 Upgrading
99 ---------
100 - 0.8.0 and 0.8.1 shipped with a bug that was setting the
101 replicate_on_write option for counter column families to false (this
102 option has no effect on non-counter column family). This is an unsafe
103 default and 0.8.2 correct this, the default for replicate_on_write is
104 now true. It is advised to update your counter column family definitions
105 if replicate_on_write was uncorrectly set to false (before or after
106 upgrade).
107
108
f491265 @jbellis merge from 0.8
jbellis authored
109 0.8.1
110 =====
111
112 Upgrading
113 ---------
114 - 0.8.1 is backwards compatible with 0.8, upgrade can be achieved by a
115 simple rolling restart.
116 - If upgrading for earlier version (0.7), please refer to the 0.8 section
117 for instructions.
118
119 Features
120 --------
121 - Numerous additions/improvements to CQL (support for counters, TTL, batch
122 inserts/deletes, index dropping, ...).
123 - Add two new AbstractTypes (comparator) to support compound keys
124 (CompositeType and DynamicCompositeType), as well as a ReverseType to
125 reverse the order of any existing comparator.
126 - New option to bypass the commit log on some keyspaces (for advanced
127 users).
128
129 Tools
130 -----
131 - Add new data bulk loading utility (sstableloader).
132
133
5830799 @jbellis merge from 0.8
jbellis authored
134 0.8
135 ===
0d3e16b update NEWS.txt to document avro package changes
Eric Evans authored
136
137 Upgrading
138 ---------
1adcd1c @jbellis merge from 0.8
jbellis authored
139 - Upgrading from version 0.7.1 or later can be done with a rolling
140 restart, one node at a time. You do not need to bring down the
921f499 @jbellis merge from 0.8
jbellis authored
141 whole cluster at once.
835669a @jbellis merge from 0.8
jbellis authored
142 - Running nodetool drain before shutting down the 0.7 node is
143 recommended but not required. (Skipping this will result in
144 replay of entire commitlog, so it will take longer to restart but
145 is otherwise harmless.)
921f499 @jbellis merge from 0.8
jbellis authored
146 - 0.8 is fully API-compatible with 0.7. You can continue
147 to use your 0.7 clients.
1adcd1c @jbellis merge from 0.8
jbellis authored
148 - Avro record classes used in map/reduce and Hadoop streaming code have
be67fda @tjake merge from 0.8
tjake authored
149 been removed. Map/reduce can be switched to Thrift by changing
150 org.apache.cassandra.avro in import statements to
151 org.apache.cassandra.thrift (no class names change). Streaming support
152 has been removed for the time being.
1adcd1c @jbellis merge from 0.8
jbellis authored
153 - The loadbalance command has been removed from nodetool. For similar
154 behavior, decommission then rebootstrap with empty initial_token.
70c0ed8 @jbellis merge from 0.8
jbellis authored
155 - Thrift unframed mode has been removed.
2ec8621 @tjake merge from 0.8
tjake authored
156 - The addition of key_validation_class means the cli will assume keys
157 are bytes, instead of strings, in the absence of other information.
158 See http://wiki.apache.org/cassandra/FAQ#cli_keys for more details.
159
0d3e16b update NEWS.txt to document avro package changes
Eric Evans authored
160
1adcd1c @jbellis merge from 0.8
jbellis authored
161 Features
162 --------
163 - added CQL client API and JDBC/DBAPI2-compliant drivers for Java and
164 Python, respectively (see: drivers/ subdirectory and doc/cql)
165 - added distributed Counters feature;
166 see http://wiki.apache.org/cassandra/Counters
167 - optional intranode encryption; see comments around 'encryption_options'
168 in cassandra.yaml
169 - compaction multithreading and rate-limiting; see
170 'compaction_multithreading' and 'compaction_throughput_mb_per_sec' in
171 cassandra.yaml
bec125b @jbellis merge from 0.8
jbellis authored
172 - cassandra will limit total memtable memory usage to 1/3 of the heap
173 by default. This can be ajusted or disabled with the
174 memtable_total_space_in_mb option. The old per-ColumnFamily
175 throughput, operations, and age settings are still respected but
176 will be removed in a future major release once we are satisfied that
177 memtable_total_space_in_mb works adequately.
5830799 @jbellis merge from 0.8
jbellis authored
178
0427c83 @jbellis merge from 0.8
jbellis authored
179 Tools
180 -----
181 - stress and py_stress moved from contrib/ to tools/
182 - clustertool was removed (see
183 https://issues.apache.org/jira/browse/CASSANDRA-2607 for examples
184 of how to script nodetool across the cluster instead)
185
018a862 update release notes and changelog for #1933
Eric Evans authored
186 Other
187 -----
8897ad5 @jbellis merge from 0.8
jbellis authored
188 - In the past, sstable2json would write column names and values as
189 hex strings, and now creates human readable values based on the
190 comparator/validator. As a result, JSON dumps created with
191 older versions of sstable2json are no longer compatible with
192 json2sstable, and imports must be made with a configuration that
193 is identical to the export.
194 - manually-forced compactions ("nodetool compact") will do nothing
195 if only a single SSTable remains for a ColumnFamily. To force it
196 to compact that anyway (which will free up space if there are
197 a lot of expired tombstones), use the new forceUserDefinedCompaction
198 JMX method on CompactionManager.
be67fda @tjake merge from 0.8
tjake authored
199 - most of contrib/ (which was not part of the binary releases)
200 has been moved either to examples/ or tools/. We plan to move the
201 rest for 0.8.1.
018a862 update release notes and changelog for #1933
Eric Evans authored
202
7fb8170 document new default JMX port
Eric Evans authored
203 JMX
204 ---
205 - By default, JMX now listens on port 7199.
206
0d3e16b update NEWS.txt to document avro package changes
Eric Evans authored
207
c330ee7 @pcmanus Update files in preparation for 0.7.6 release
pcmanus authored
208 0.7.6
209 =====
210
211 Upgrading
212 ---------
213 - Nothing specific to 0.7.6, but see 0.7.3 Upgrading if upgrading
214 from earlier than 0.7.1.
215
216
979ce41 @jbellis fix migration race vs flush
jbellis authored
217 0.7.5
218 =====
219
220 Upgrading
221 ---------
3c988f7 @pcmanus Update NEWS.txt for 0.7.5 release
pcmanus authored
222 - Nothing specific to 0.7.5, but see 0.7.3 Upgrading if upgrading
979ce41 @jbellis fix migration race vs flush
jbellis authored
223 from earlier than 0.7.1.
224
225 Changes
226 -------
227 - system_update_column_family no longer snapshots before applying
228 the schema change. (_update_keyspace never did. _drop_keyspace
229 and _drop_column_family continue to snapshot.)
3c988f7 @pcmanus Update NEWS.txt for 0.7.5 release
pcmanus authored
230 - added memtable_flush_queue_size option to cassandra.yaml to
231 avoid blocking writes when multiple column families (or a colum
232 family with indexes) are flushed at the same time.
233 - allow overriding initial_token, storage_port and rpc_port using
234 system properties
979ce41 @jbellis fix migration race vs flush
jbellis authored
235
236
29898b8 @jbellis update NEWS for 0.7.4
jbellis authored
237 0.7.4
238 =====
239
240 Upgrading
241 ---------
242 - Nothing specific to 0.7.4, but see 0.7.3 Upgrading if upgrading
243 from earlier than 0.7.1.
244
245 Features
246 --------
247 - Output to Pig is now supported as well as input
248
249
21bcf93 @jbellis add nodetool scrub to NEWS
jbellis authored
250 0.7.3
251 =====
252
253 Upgrading
254 ---------
255 - 0.7.1 and 0.7.2 shipped with a bug that caused incorrect row-level
256 bloom filters to be generated when compacting sstables generated
257 with earlier versions. This would manifest in IOExceptions during
258 column name-based queries. 0.7.3 provides "nodetool scrub" to
259 rebuild sstables with correct bloom filters, with no data lost.
260 (If your cluster was never on 0.7.0 or earlier, you don't have to
261 worry about this.) Note that nodetool scrub will snapshot your
262 data files before rebuilding, just in case.
263
264
13bf76c @jbellis enable #1530 only after cluster is all on 0.7.1
jbellis authored
265 0.7.1
266 =====
267
5e18204 @jbellis add flush_largest_memtables_at and reduce_cache_sizes_at options
jbellis authored
268 Upgrading
269 ---------
13bf76c @jbellis enable #1530 only after cluster is all on 0.7.1
jbellis authored
270 - 0.7.1 is completely backwards compatible with 0.7.0. Just restart
271 each node with the new version, one at a time. (The cluster does
272 not all need to be upgraded simultaneously.)
273
274 Features
275 --------
5e18204 @jbellis add flush_largest_memtables_at and reduce_cache_sizes_at options
jbellis authored
276 - added flush_largest_memtables_at and reduce_cache_sizes_at options
277 to cassandra.yaml as an escape valve for memory pressure
278 - added option to specify -Dcassandra.join_ring=false on startup
279 to allow "warm spare" nodes or performing JMX maintenance before
280 joining the ring
281
282 Performance
283 -----------
284 - Disk writes and sequential scans avoid polluting page cache
285 (requires JNA to be enabled)
286 - Cassandra performs writes efficiently across datacenters by
13bf76c @jbellis enable #1530 only after cluster is all on 0.7.1
jbellis authored
287 sending a single copy of the mutation and having the recipient
288 forward that to other replicas in its datacenter.
5e18204 @jbellis add flush_largest_memtables_at and reduce_cache_sizes_at options
jbellis authored
289 - Improved network buffering
290 - Reduced lock contention on memtable flush
291 - Optimized supercolumn deserialization
292 - Zero-copy reads from mmapped sstable files
293 - Explicitly set higher JVM new generation size
294 - Reduced i/o contention during saving of caches
13bf76c @jbellis enable #1530 only after cluster is all on 0.7.1
jbellis authored
295
296
dbd6daa release note mentioning renamed directives
Eric Evans authored
297 0.7.0
298 =====
299
602cf75 @jbellis update NEWS, configuration for byte[] keys change. patch by Stu Hood; r...
jbellis authored
300 Features
301 --------
e42192e @jbellis update NEWS for 0.7. patch by gdusbabek and jbellis for CASSANDRA-1395
jbellis authored
302 - Secondary indexes (indexes on column values) are now supported
303 - Row size limit increased from 2GB to 2 billion columns. rows
304 are no longer read into memory during compaction.
305 - Keyspace and ColumnFamily definitions may be added and modified live
306 - Streaming data for repair or node movement no longer requires
307 anticompaction step first
916c810 @jbellis rename RackAwareStrategy to OldNetworkTopologyStrategy, RackUnawareStrat...
jbellis authored
308 - NetworkTopologyStrategy (formerly DatacenterShardStrategy) is ready for
f1af79b spelling error (thanks Dan Di Spaltro :))
Eric Evans authored
309 use, enabling ConsistencyLevel.DCQUORUM and DCQUORUMSYNC. See comments
916c810 @jbellis rename RackAwareStrategy to OldNetworkTopologyStrategy, RackUnawareStrat...
jbellis authored
310 in `cassandra.yaml.`
e42192e @jbellis update NEWS for 0.7. patch by gdusbabek and jbellis for CASSANDRA-1395
jbellis authored
311 - Optional per-Column time-to-live field allows expiring data without
312 have to issue explicit remove commands
313 - `truncate` thrift method allows clearing an entire ColumnFamily at once
8c5069f @jbellis support for Hadoop Streaming [non-jvm map/reduce via stdin/out]. patch ...
jbellis authored
314 - Hadoop OutputFormat and Streaming [non-jvm map/reduce via stdin/out]
315 support
e42192e @jbellis update NEWS for 0.7. patch by gdusbabek and jbellis for CASSANDRA-1395
jbellis authored
316 - Up to 8x faster reads from row cache
317 - A new ByteOrderedPartitioner supports bytes keys with arbitrary content,
318 and orders keys by their byte value. This should be used in new
319 deployments instead of OrderPreservingPartitioner.
320 - Optional round-robin scheduling between keyspaces for multitenant
0da82f3 @jbellis implement keyspace round-robin scheduler. patch by Nirmal Ranganathan; ...
jbellis authored
321 clusters
e42192e @jbellis update NEWS for 0.7. patch by gdusbabek and jbellis for CASSANDRA-1395
jbellis authored
322 - Dynamic endpoint snitch mitigates the impact of impaired nodes
323 - New `IntegerType`, faster than LongType and allows integers of
324 both less and more bits than Long's 64
043bc82 document authorization changes
Eric Evans authored
325 - A revamped authentication system that decouples authorization and
326 allows finer-grained control of resources.
e42192e @jbellis update NEWS for 0.7. patch by gdusbabek and jbellis for CASSANDRA-1395
jbellis authored
327
328 Upgrading
329 ---------
330 The Thrift API has changed in incompatible ways; see below, and refer
331 to http://wiki.apache.org/cassandra/ClientOptions for a list of
332 higher-level clients that have been updated to support the 0.7 API.
333
4023c3b @jbellis derive stage from verb instead of transmitting it for each message
jbellis authored
334 The Cassandra inter-node protocol is incompatible with 0.6.x
335 releases (and with 0.7 beta1), meaning you will have to bring your
336 cluster down prior to upgrading: you cannot mix 0.6 and 0.7 nodes.
d2667f6 truncate hints if starting 0.7 for the first time. patch by gdusbabek, r...
Gary Dusbabek authored
337
338 The hints schema was changed from 0.6 to 0.7. Cassandra automatically
339 snapshots and then truncates the hints column family as part of
340 starting up 0.7 for the first time.
e42192e @jbellis update NEWS for 0.7. patch by gdusbabek and jbellis for CASSANDRA-1395
jbellis authored
341
342 Keyspace and ColumnFamily definitions are stored in the system
343 keyspace, rather than the configuration file.
344
345 The process to upgrade is:
346 1) run "nodetool drain" on _each_ 0.6 node. When drain finishes (log
347 message "Node is drained" appears), stop the process.
348 2) Convert your storage-conf.xml to the new cassandra.yaml using
7674982 CHANGES.txt and NEWS.txt update explaining the ramifications of CASSANDR...
Gary Dusbabek authored
349 "bin/config-converter".
350 3) Rename any of your keyspace or column family names that do not adhere
351 to the '^\w+' regex convention.
1e47c83 clarify wording in release notes
Eric Evans authored
352 4) Start up your cluster with the 0.7 version.
7674982 CHANGES.txt and NEWS.txt update explaining the ramifications of CASSANDR...
Gary Dusbabek authored
353 5) Initialize your Keyspace and ColumnFamily definitions using
e42192e @jbellis update NEWS for 0.7. patch by gdusbabek and jbellis for CASSANDRA-1395
jbellis authored
354 "bin/schematool <host> <jmxport> import". _You only need to do
355 this to one node_.
356
357 Thrift API
358 ----------
42fa6c0 @jbellis mention framed mode in NEWS
jbellis authored
359 - The Cassandra server now defaults to framed mode, rather than
9cf9e93 @jbellis emphasize that unframed mode is going away sooner than later
jbellis authored
360 unframed. Unframed is obsolete and will be removed in the next
361 major release.
2d9b4ba @jbellis update contrib WordCount, ClientOnlyExample for Thrift 0.5. patch by jb...
jbellis authored
362 - The Cassandra Thrift interface file has been updated for Thrift 0.5.
363 If you are compiling your own client code from the interface, you
364 will need to upgrade the Thrift compiler to match.
e42192e @jbellis update NEWS for 0.7. patch by gdusbabek and jbellis for CASSANDRA-1395
jbellis authored
365 - Row keys are now bytes: keys stored by versions prior to 0.7.0 will be
366 returned as UTF-8 encoded bytes. OrderPreservingPartitioner and
367 CollatingOrderPreservingPartitioner continue to expect that keys contain
368 UTF-8 encoded strings, but RandomPartitioner now works on any key data.
369 - keyspace parameters have been replaced with the per-connection
370 set_keyspace method.
371 - The return type for login() is now AccessLevel.
372 - The get_string_property() method has been removed.
373 - The get_string_list_property() method has been removed.
602cf75 @jbellis update NEWS, configuration for byte[] keys change. patch by Stu Hood; r...
jbellis authored
374
dbd6daa release note mentioning renamed directives
Eric Evans authored
375 Configuraton
376 ------------
0438494 @jbellis convert cassandra.xml to cassandra.yaml. patch by Todd Blose and Stu Ho...
jbellis authored
377 - Configuration file renamed to cassandra.yaml and log4j.properties to
dd787f7 @jbellis simplify configuration file loading; rename to cassandra.xml
jbellis authored
378 log4j-server.properties
0f89876 @jbellis forward-port CASSANDRA-1481. originally thought it was a non-issue in 0...
jbellis authored
379 - PropertyFileSnitch configuration file renamed to
380 cassandra-topology.properties
dbd6daa release note mentioning renamed directives
Eric Evans authored
381 - The ThriftAddress and ThriftPort directives have been renamed to
382 RPCAddress and RPCPort respectively.
1fe2221 @jbellis update docs for CASSANDRA-994
jbellis authored
383 - EndPointSnitch was renamed to RackInferringSnitch. A new SimpleSnitch
384 has been added.
e42192e @jbellis update NEWS for 0.7. patch by gdusbabek and jbellis for CASSANDRA-1395
jbellis authored
385 - RackUnawareStrategy and RackAwareStrategy have been renamed to
386 SimpleStrategy and OldNetworkTopologyStrategy, respectively.
395c3ca @jbellis make GCGraceSeconds configurable per-columnfamily. patch by Jon Hermes;...
jbellis authored
387 - RowWarningThresholdInMB replaced with in_memory_compaction_limit_in_mb
388 - GCGraceSeconds is now per-ColumnFamily instead of global
7674982 CHANGES.txt and NEWS.txt update explaining the ramifications of CASSANDR...
Gary Dusbabek authored
389 - Keyspace and column family names that do not confirm to a '^\w+' regex
390 are considered illegal.
801a26f @driftx Mention schematool in the config section of NEWS.txt
driftx authored
391 - Keyspace and column family definitions will need to be loaded via
392 "bin/schematool <host> <jmxport> import". _You only need to do this to
393 one node_.
043bc82 document authorization changes
Eric Evans authored
394 - In addition to an authenticator, an authority must be configured as
395 well. Users of SimpleAuthenticator should use SimpleAuthority for this
396 value (the default is AllowAllAuthority, which corresponds with
397 AllowAllAuthenticator).
398 - The format of access.properties has changed, see the sample configuration
399 conf/access.properties for documentation on the new format.
801a26f @driftx Mention schematool in the config section of NEWS.txt
driftx authored
400
dbd6daa release note mentioning renamed directives
Eric Evans authored
401
2fa9381 @jbellis move StreamingService from o.a.c.streaming to o.a.c.service. patch by j...
jbellis authored
402 JMX
403 ---
404 - StreamingService moved from o.a.c.streaming to o.a.c.service
8c0c3eb @jbellis rename GMFD to GOSSIP_STAGE. patch by Rob Coli; reviewed by jbellis for...
jbellis authored
405 - GMFD renamed to GOSSIP_STAGE
e3f05ad @jbellis {Min,Mean,Max}RowCompactedSize renamed to {Min,Mean,Max}RowSize. patch ...
jbellis authored
406 - {Min,Mean,Max}RowCompactedSize renamed to {Min,Mean,Max}RowSize
407 since it no longer has to wait til compaction to be computed
2fa9381 @jbellis move StreamingService from o.a.c.streaming to o.a.c.service. patch by j...
jbellis authored
408
f8d43ed Remove references to -dev version of pig, add example script, use compar...
Johan Oskarsson authored
409 Other
410 -----
411 - If extending AbstractType, make sure you follow the singleton pattern
e42192e @jbellis update NEWS for 0.7. patch by gdusbabek and jbellis for CASSANDRA-1395
jbellis authored
412 followed by Cassandra core AbstractType classes: provide a public
413 static final variable called 'instance'.
a58c7fd document changes made in CASSANDRA-900
Eric Evans authored
414
3001070 @jbellis merge from 0.6
jbellis authored
415
43e6b7f @jbellis merge from 0.6
jbellis authored
416 0.6.6
417 =====
418
419 Upgrading
420 ---------
421 - As part of the cache-saving feature, a third directory
422 (along with data and commitlog) has been added to the config
423 file. You will need to set and create this directory
424 when restarting your node into 0.6.6.
425
426
3001070 @jbellis merge from 0.6
jbellis authored
427 0.6.1
428 =====
429
430 Upgrading
431 ---------
432 - We try to keep minor versions 100% compatible (data format,
433 commitlog format, network format) within the major series, but
a60fe7c @driftx Fix minor grammar mistake in NEWS.txt
driftx authored
434 we introduced a network-level incompatibility in 0.6.1.
3001070 @jbellis merge from 0.6
jbellis authored
435 Thus, if you are upgrading from 0.6.0 to any higher version
436 (0.6.1, 0.6.2, etc.) then you will need to restart your entire
437 cluster with the new version, instead of being able to do a
438 rolling restart.
439
440
b074955 update release notes for config changes
Eric Evans authored
441 0.6.0
442 =====
3bf9c81 @jbellis update CHANGELOG, NEWS, and thrift version string. patch by jbellis
jbellis authored
443
444 Features
445 --------
713710b Merge with 0.6 branch
Johan Oskarsson authored
446 - row caching: configure with the RowsCached attribute in
447 ColumnFamily definition
2b69c40 @jbellis merge from 0.6
jbellis authored
448 - Hadoop map/reduce support: see contrib/word_count for an example
713710b Merge with 0.6 branch
Johan Oskarsson authored
449 - experimental authentication support, described under
450 Authenticator in storage.conf
3bf9c81 @jbellis update CHANGELOG, NEWS, and thrift version string. patch by jbellis
jbellis authored
451
452 Configuraton
453 ------------
454 - MemtableSizeInMB has been replaced by MemtableThroughputInMB which
455 triggers a memtable flush when the specified amount of data has
456 been written, including overwrites.
457 - MemtableObjectCountInMillions has been replaced by the
458 MemtableOperationsInMillions directive which causes a memtable flush
459 to occur after the specified number of operations.
460 - Like MemtableSizeInMB, BinaryMemtableSizeInMB has been replaced by
461 BinaryMemtableThroughputInMB.
03278af merge with 0.6 branch
Eric Evans authored
462 - Replication factor is now per-keyspace, rather than global.
1de6863 @jbellis merge from 0.6
jbellis authored
463 - KeysCachedFraction is deprecated in favor of KeysCached
8f1376b @jbellis merge from 0.6
jbellis authored
464 - RowWarningThresholdInMB added, to warn before very large rows
465 get big enough to threaten node stability
3bf9c81 @jbellis update CHANGELOG, NEWS, and thrift version string. patch by jbellis
jbellis authored
466
467 Thrift API
468 ----------
713710b Merge with 0.6 branch
Johan Oskarsson authored
469 - removed deprecated get_key_range method
470 - added batch_mutate meethod
471 - deprecated multiget and batch_insert methods in favor of
472 multiget_slice and batch_mutate, respectively
03278af merge with 0.6 branch
Eric Evans authored
473 - added ConsistencyLevel.ANY, for when you want write
474 availability even when it may not be readable immediately.
475 Unlike CL.ZERO, though, it will throw an exception if
476 it cannot be written *somewhere*.
3bf9c81 @jbellis update CHANGELOG, NEWS, and thrift version string. patch by jbellis
jbellis authored
477
478 JMX metrics
479 -----------
480 - read and write statistics are reported as lifetime totals,
713710b Merge with 0.6 branch
Johan Oskarsson authored
481 instead of averages over the last minute. average-since-last
482 requested are also available for convenience.
92b9e41 @jbellis move commitlog jmx to cassandra.db. patch by Robert Coli; reviewed by j...
jbellis authored
483 - cache hit rate statistics are now available from JMX under
484 org.apache.cassandra.db.Caches
3bf9c81 @jbellis update CHANGELOG, NEWS, and thrift version string. patch by jbellis
jbellis authored
485 - compaction JMX metrics are moved to
486 org.apache.cassandra.db.CompactionManager. PendingTasks is now
487 a much better estimate of compactions remaining, and the
488 progress of the current compaction has been added.
92b9e41 @jbellis move commitlog jmx to cassandra.db. patch by Robert Coli; reviewed by j...
jbellis authored
489 - commitlog JMX metrics are moved to org.apache.cassandra.db.Commitlog
78cbce1 @jbellis merge from 0.6
jbellis authored
490 - progress of data streaming during bootstrap, loadbalance, or other
491 data migration, is available under
492 org.apache.cassandra.streaming.StreamingService.
493 See http://wiki.apache.org/cassandra/Streaming for details.
3bf9c81 @jbellis update CHANGELOG, NEWS, and thrift version string. patch by jbellis
jbellis authored
494
8b61388 merge with 0.6 branch
Eric Evans authored
495 Installation/Upgrade
496 --------------------
713710b Merge with 0.6 branch
Johan Oskarsson authored
497 - 0.6 network traffic is not compatible with earlier versions. You
498 will need to shut down all your nodes at once, upgrade, then restart.
499
8b61388 merge with 0.6 branch
Eric Evans authored
500
b074955 update release notes for config changes
Eric Evans authored
501
49197c9 @jbellis add NEWS.txt release notes
jbellis authored
502 0.5.0
503 =====
713710b Merge with 0.6 branch
Johan Oskarsson authored
504
49197c9 @jbellis add NEWS.txt release notes
jbellis authored
505 0. The commitlog format has changed (but sstable format has not).
506 When upgrading from 0.4, empty the commitlog either by running
507 bin/nodeprobe flush on each machine and waiting for the flush to finish,
508 or simply remove the commitlog directory if you only have test data.
509 (If more writes come in after the flush command, starting 0.5 will error
510 out; if that happens, just go back to 0.4 and flush again.)
d4d8381 @jbellis merge from 0.5 branch
jbellis authored
511 The format changed twice: from 0.4 to beta1, and from beta2 to RC1.
49197c9 @jbellis add NEWS.txt release notes
jbellis authored
512
c10813c @jbellis merge from 0.5
jbellis authored
513 .5 The gossip protocol has changed, meaning 0.5 nodes cannot coexist
514 in a cluster of 0.4 nodes or vice versa; you must upgrade your
515 whole cluster at the same time.
516
3e574a9 @jbellis only GC when compacting all sstables, to avoid situations where the data...
jbellis authored
517 1. Bootstrap, move, load balancing, and active repair have been added.
dc519a1 merge with 0.5 branch
Eric Evans authored
518 See http://wiki.apache.org/cassandra/Operations. When upgrading
519 from 0.4, leave autobootstrap set to false for the first restart
520 of your old nodes.
49197c9 @jbellis add NEWS.txt release notes
jbellis authored
521
522 2. Performance improvements across the board, especially on the write
523 path (over 100% improvement in stress.py throughput).
524
525 3. Configuration:
526 - Added "comment" field to ColumnFamily definition.
527 - Added MemtableFlushAfterMinutes, a global replacement for the
528 old per-CF FlushPeriodInMinutes setting
529 - Key cache settings
530
531 4. Thrift:
532 - Added get_range_slice, deprecating get_key_range
533
534
713710b Merge with 0.6 branch
Johan Oskarsson authored
535
49197c9 @jbellis add NEWS.txt release notes
jbellis authored
536 0.4.2
537 =====
713710b Merge with 0.6 branch
Johan Oskarsson authored
538
49197c9 @jbellis add NEWS.txt release notes
jbellis authored
539 1. Improve default garbage collector options significantly --
540 throughput will be 30% higher or more.
541
542
713710b Merge with 0.6 branch
Johan Oskarsson authored
543
49197c9 @jbellis add NEWS.txt release notes
jbellis authored
544 0.4.1
545 =====
713710b Merge with 0.6 branch
Johan Oskarsson authored
546
49197c9 @jbellis add NEWS.txt release notes
jbellis authored
547 1. SnapshotBeforeCompaction configuration option allows snapshotting
548 before each compaction, which allows rolling back to any version
549 of the data.
550
551
713710b Merge with 0.6 branch
Johan Oskarsson authored
552
49197c9 @jbellis add NEWS.txt release notes
jbellis authored
553 0.4.0
554 =====
713710b Merge with 0.6 branch
Johan Oskarsson authored
555
49197c9 @jbellis add NEWS.txt release notes
jbellis authored
556 1. On-disk data format has changed to allow billions of keys/rows per
557 node instead of only millions. The new format is incompatible with 0.3;
558 see 0.3 notes below for how to import data from a 0.3 install.
559
560 2. Cassandra now supports multiple keyspaces. Typically you will have
561 one keyspace per application, allowing applications to be able to
562 create and modify ColumnFamilies at will without worrying about
563 collisions with others in the same cluster.
564
565 3. Many Thrift API changes and documentation. See
566 http://wiki.apache.org/cassandra/API
567
568 4. Removed the web interface in favor of JMX and bin/nodeprobe, which
569 has significantly enhanced functionality.
570
571 5. Renamed configuration "<Table>" to "<Keyspace>".
572
573 6. Added commitlog fsync; see "<CommitLogSync>" in configuration.
574
575
713710b Merge with 0.6 branch
Johan Oskarsson authored
576
49197c9 @jbellis add NEWS.txt release notes
jbellis authored
577 0.3.0
578 =====
713710b Merge with 0.6 branch
Johan Oskarsson authored
579
49197c9 @jbellis add NEWS.txt release notes
jbellis authored
580 1. With enough and large enough keys in a ColumnFamily, Cassandra will
581 run out of memory trying to perform compactions (data file merges).
582 The size of what is stored in memory is (S + 16) * (N + M) where S
583 is the size of the key (usually 2 bytes per character), N is the
584 number of keys and M, is the map overhead (which can be guestimated
585 at around 32 bytes per key).
586 So, if you have 10-character keys and 1GB of headroom in your heap
587 space for compaction, you can expect to store about 17M keys
588 before running into problems.
589 See https://issues.apache.org/jira/browse/CASSANDRA-208
590
591 2. Because fixing #1 requires a data file format change, 0.4 will not
592 be binary-compatible with 0.3 data files. A client-side upgrade
593 can be done relatively easily with the following algorithm:
594 for key in old_client.get_key_range(everything):
595 columns = old_client.get_slice or get_slice_super(key, all columns)
596 new_client.batch_insert or batch_insert_super(key, columns)
597 The inner loop can be trivially parallelized for speed.
598
599 3. Commitlog does not fsync before reporting a write successful.
600 Using blocking writes mitigates this to some degree, since all
601 nodes that were part of the write quorum would have to fail
602 before sync for data to be lost.
603 See https://issues.apache.org/jira/browse/CASSANDRA-182
604
605 Additionally, row size (that is, all the data associated with a single
606 key in a given ColumnFamily) is limited by available memory, because
607 compaction deserializes each row before merging.
608
609 See https://issues.apache.org/jira/browse/CASSANDRA-16
4e14314 testing svn access
Gary Dusbabek authored
610
Something went wrong with that request. Please try again.