Crate 0.55.0 发布数据存储系统

Crate 0.55.0 发布。主要更新内容如下：

– BREAKING: Units are now required for time or byte based settings in the
`crate.yml` configuration file. Nodes won’t start otherwise.
– BREAKING: The `gateway.local.sync` setting has been renamed to
`translog.sync_interval`.
– BREAKING: It is no longer possible to specify the location of the config
file with the `CONF_FILE` environment variable, or the `-Des.config`,

`-Des.default.config`, or `-Delasticsearch.config` parameters.
The config file must now be named `crate.yml` and must be located in the
default `config/` directory, in the directory specified in the `CONF_DIR`

environmental variable, or the one specified in the `-Des.path.home`

parameter.
– BREAKING: Crate will die now on all plugin loading errors.
– BREAKING: Column names may not contain the `.` character anymore.
– BREAKING: Tables are not refreshed automatically anymore before and after
a `DELETE-BY-QUERY` operation is applied.
– Upgraded Elasticsearch to 2.3.3
– Updated crate-admin to 0.18.0 which contains following changes:
– Cluster check include now clickable links
– Load additional links in menu dynamically from news feed
– Fixed an issue which caused rendering errors in the node list if the cluster
contains multiple nodes with the same name

note: changes the url to the detail view of single nodes
– Added UTM tokens to links that link to the Crate.IO blog
– Updated news feed URL
– Label the master node in the cluster node list
– Fixed a bug which prevented complete deletion of tables in custom
schemas.
– Improved the `KILL` behaviour for `COPY`, `UPDATE` and `DELETE` statements.
– Fixed a `ClassCastException` that occurred if the left side of a `IN`
predicate was a function.
– Added links to Cluster check documentation.
– Removed the `creationTime` property from `SQL(Bulk)Request` and changed the
`duration` property on `SQL(Bulk)Response` to return the server-side

duration instead of the full round-trip duration.
– The default number of shards is calculated dynamically upon the table
creation.
– Deprecated camelCase names on `CREATE ANALYZER`.
– `+HeapDumpOnOutOfMemoryError` is no longer set by default.
It can still be set using the JAVA_OPTS environment variable.
– Made sure orphaned contexts are cleaned up eventually after a coordinator
node dies.
– Load “crate.yml“ explicitly and do not allow any other extension than
“yml“ for the configuration file.
– Restrict creation of tables having the primary key constraint within
an array column type or its children.
– Renamed all scalar function names to lowercase (CamelCase names
were transformed to lowercase using `_` as word boundary)
– Fixed a NPE at the `substr` scalar function when using column
references as from/to arguments containing null values.
– NOTE: Upgrading to this Crate version is only supported from >= 0.52
– Removed the `bulk.partition_creation_timeout` setting.
– Removed internal migration code that ensured schema backward compatibility
to Crate versions <= 0.45
– Improved the speed of global aggregation queries which involve more than 1
shard per node.
– The log level of loggers can now be set at runtime via the “SET“ statement.
– Improved validation of geo point data type; now it is not possible any more
to insert invalid latitude or longitude in array literals.
– Nodes can be configured now to allow only SQL statements resulting in read operations using new setting “node.sql.read_only“.
– Improved bulk insert/update and copy-from performance when replicas
are configured.
– Fixed planner issues that could cause INSERT INTO with a JOIN in the
sub-query to fail or get stuck.
– Removed the “jobs.keep_alive_timeout“ setting and all related logic.
This means it is no longer possible to enable automatic job termination.
– Added a “node_filters“ option to “COPY FROM“.
– Add support for “COPY FROM“ to “EXPLAIN“.
– The “COPY FROM“ statement now also accepts an array of strings as URI.
– Added support for schemes supported by the used JVM’s `URL` implementation
for “COPY FROM“. These usually include“http“, “https“, “ftp“ and

“jar“.
– Added support for the “unnest“ table function.
– Implemented cluster check for installed java version.
– removed wrong behaving CrateClient feature which allowed to load
client settings from a configuration file.
– Crate will now recognize settings defined by crate plugins.
– Implemented the EXPLAIN command for SELECT statements.
– Removed the limitation that the WHERE clause couldn’t contain
certain expressions in `DELETE` statements
– Exposed setting the default schema via HTTP header.
– Added `lower“ and `upper“ string functions.