Drill can run in 2 modes - Embedded and Distributed. In Distributed Mode it uses ZK for keeping track of all the Drillbits in a Drill cluster. When connecting to a a distributed cluster you can either connect direct to a single drill or use ZK to get a list if all drillbits and then get connected to a drillbit in the ZK list to manage the session. Using ZK is preferred as it allows clients to connect to the cluster and any one of the drillbits to distribute the session and foreman load, if one drillbit fails clients can still connect to the cluster.
In Embedded mode ZK is not used to maintain a list of drillbits, and typically ZK is not configured (not needed). In this case you basically have to connect directly to the drillbit. You tried to connect to ZK in embedded mode when no ZK was probably available. An alternative is to configure a 1 node cluster with a drillbit and ZK, and configure Drill in Distributed mode. This allows you to work with Drill as if it is cluster although a single node. The MapR Drill sandbox is a greta example of this and very handy. Also note that when using Distributed mode and ZK it is always a good idea to make sure hostnames of ZK and the Drillbits can be resolved by the client, not just the IPs. Also good to validate that a client can reach all the nodes with drillbits and ZK on the network, to avoid a client session being assigned to a drillbit it can’t reach. —Andries > On Nov 3, 2015, at 2:41 AM, Martin Senne <[email protected]> wrote: > > Hi Andries, hi all, > > great stuff, > > drillbit=localhost instead of > zk=local > > did the trick. > > Could you elaborate a little more, on what the difference is or where to > retrieve additional information / documentation? > > Thx > > Martin > > > > 2015-11-02 19:26 GMT+01:00 Andries Engelbrecht <[email protected] > <mailto:[email protected]>>: > Looks like you are using Drill in embedded mode. > Instead of trying to connect to ZK (not available in embedded mode) rather > connect to the drillbit directly. > > Example below > sqlline -u jdbc:drill:drillbit=localhost > > Basically replace zk=localhost with drillbit=localhost > > See if that works for you. > > —Andries > > > On Nov 2, 2015, at 9:48 AM, Martin Senne <[email protected] > > <mailto:[email protected]>> wrote: > > > > Hi all, > > > > Situation > > > > following the details on > > https://drill.apache.org/docs/using-the-jdbc-driver/ > > <https://drill.apache.org/docs/using-the-jdbc-driver/> > > I'm trying to connect > > > > - with a simple JDBC Client (based on the code on bottom of > > "using-the-jdbc-driver"-page) > > - using the Drill JDBC Driver( > > *./jars/jdbc-driver/drill-jdbc-all-1.2.0.jar* ) in classpath > > - to a locally running embedded drill instance. ( started via > > ./bin/drill-embedded ) > > - with connection url:* jdbc:drill:schema=P3.public;zk=local* > > - Drill Storage P3 is of type "jdbc" and connects to a PostgreSQL DB. > > (Remark: Storage P3 is working properly and can be queried in embedded > > Drill shell as expected) > > > > Problem > > > > On startup of the simple JDBC Client, I encounter: > > > > 17:36:44 DEBUG FormatPluginConfigBase: Found 0 format plugin configuration > > classes: . > > 17:36:44 DEBUG DrillConfig: DrillConfig object initialized. > > [error] (run-main-0) java.lang.NoClassDefFoundError: > > oadd/org/apache/drill/exec/store/sys/PStoreProvider > > java.lang.NoClassDefFoundError: > > *oadd/org/apache/drill/exec/store/sys/PStoreProvider* > > at > > org.apache.drill.jdbc.impl.DrillConnectionImpl.<init>(DrillConnectionImpl.java:100) > > at > > org.apache.drill.jdbc.impl.DrillJdbc41Factory.newDrillConnection(DrillJdbc41Factory.java:66) > > at > > org.apache.drill.jdbc.impl.DrillFactory.newConnection(DrillFactory.java:69) > > at > > oadd.net.hydromatic.avatica.UnregisteredDriver.connect(UnregisteredDriver.java:126) > > at org.apache.drill.jdbc.Driver.connect(Driver.java:72) > > at java.sql.DriverManager.getConnection(DriverManager.java:664) > > at java.sql.DriverManager.getConnection(DriverManager.java:247) > > > > Analysis > > > > - Opening drill-jdbc-all-1.2.0 .jar the path > > /oadd/org/apache/drill/exec/... is missing in the jar. > > - BTW: The same holds for a freshly built drill-1.3.0-snapshot > > > > Questions > > > > - Is it reasonable (and intented) to connect to an embedded drill > > instance via the JDBC client, like I do? If not, what are the > > alternatives? > > - Is it a bug (or maybe unintended), that within ./*exec/jdbc-all/pom.xml > > *in the <filter>-section of the shade plugin, "exec/store" (line 408) is > > filtered? > > <exclude>org/apache/drill/exec/store/**</exclude> > > > > Thx for help in advance, > > > > Martin > >
