Hi Hema,
Welcome to Drill! Drill's primary use case is to run distributed queries across
multiple nodes. For that, Drill generally runs as a stand-alone process,
perhaps launched by YARN, Docker, K8s or your own scripts. Although Drill runs
in embedded mode, this is mostly to provide a "quick start" when playing with
sample and local data in Sqlline.
Take a look at the file $DRILL_HOME/bin/embedded. Just use the following JDBC
connection string: jdbc:drill:zk=local. All of Drill's jars must be on your
class path. Drill also must have visibility to its config files.
You may find, however, that Drill is quite heavy-weight when used as a library,
since this is not the main use case. For example:
* Drill will run in your root class loader. This means that Drill's many
dependencies may conflict with the versions you use.
* Drill uses storage plugin configurations to access data. In server mode,
Drill stores this information in ZK so it is available across runs. In embedded
mode, the data is stored in /tmp (which can be changed.)
* Drill will use a large amount of memory to run queries: the default is 4 GB.
Queries will run across multiple threads and make heavy use of CPU (if you have
large amounts of data.)
* If you plan to run more than one of your applications on a single node, you
may end up with port conflicts as Drill will bind to its four ports, even in
embedded mode.
While you can make an embedded Drill work, you might find you are better off
running a Drill process separately from your application.
Thanks,
- Paul
On Tuesday, February 4, 2020, 10:22:41 PM PST, Hema T <[email protected]>
wrote:
Hello,
I am working on a use-case to use drill jdbc features. The ask is on
launching a drillbit programmatically within a Java prog execution without
having it run separately in embedded/distributed mode. Any Java API
references/examples /suggestions will be helpful.
Thanks
