paul-rogers commented on a change in pull request #2060: DRILL-7706: Implement
Drill RDBMS Metastore for Tables component
URL: https://github.com/apache/drill/pull/2060#discussion_r410823535
##########
File path: metastore/rdbms-metastore/README.md
##########
@@ -0,0 +1,140 @@
+# RDBMS Metastore
+
+RDBMS Metastore implementation allows to store Drill Metastore metadata in
configured RDBMS.
+
+## Configuration
+
+Currently, RDBMS Metastore is not default Drill Metastore implementation.
+To enable RDBMS Metastore create `drill-metastore-override.conf` and indicate
RDBMS Metastore class:
+
+```
+drill.metastore: {
+ implementation.class: "org.apache.drill.metastore.rdbms.RdbmsMetastore"
+}
+```
+
+### Connection properties
+
+Data source connection properties allows to indicate how to connect to Drill
Metastore database.
+
+`drill.metastore.rdbms.data_source.driver` - driver class name. Required.
+Note, driver must be included into Drill classpath prior to start up for all
databases except of SQLite.
+
+`drill.metastore.rdbms.data_source.url` - connection url. Required.
+
+`drill.metastore.rdbms.data_source.username` - database user on whose behalf
the connection is
+being made. Optional, if database does not require user to connect.
+
+`drill.metastore.rdbms.data_source.password` - database user's password.
+Optional, if database does not require user's password to connect.
+
+`drill.metastore.rdbms.data_source.properties` - specifies properties which
will be used
+during data source creation. See list of available [Hikari
properties](https://github.com/brettwooldridge/HikariCP)
+for more details.
+
+### Default configuration
+
+Out of the box, Drill RDBMS Metastore is configured to use embedded file
system based SQLite database.
+It will be created locally in user's home directory under
`${drill.exec.zk.root}"/metastore` location.
+
+Default setup can be used only in Drill embedded mode.
+If SQLite setup will be used in distributed mode, each drillbit will have it's
own SQLite instance
+which will lead to bogus results during queries execution.
+In distributed mode, database instance must be accessible for all drillbits.
Review comment:
SQLite is an embedded database; is not distributed. SQLite is good for
trying out the feature, for testing, for a running Drill in embedded mode, and
perhaps for a single-node Drill "cluster". If should not be used in a
multi-node cluster. Each Drillbit will have its own version of the truth and
behavior will be undefined and incorrect.
((Question: can we detect that more than one Drillbit is running and fail or
log an error if using Sqllite?))
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
