[
https://issues.apache.org/jira/browse/DRILL-5015?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15653515#comment-15653515
]
ASF GitHub Bot commented on DRILL-5015:
---------------------------------------
Github user sohami commented on a diff in the pull request:
https://github.com/apache/drill/pull/648#discussion_r87351685
--- Diff:
exec/java-exec/src/main/java/org/apache/drill/exec/client/DrillClient.java ---
@@ -223,19 +223,65 @@ public void connect(Properties props) throws
RpcException {
connect(null, props);
}
+ /**
+ * Function to populate the endpointList with information of all the
drillbits
+ * provided in the connection string by client
+ * @param endpointList - ArrayList of DrillbitEndpoints
+ * @param drillbits - One or more drillbit ip[:port] provided in
connection string
+ */
+ public void populateEndpointsList(ArrayList<DrillbitEndpoint>
endpointList, String drillbits){
+
+ // If no information about drillbits is provided then just return
empty list.
+ if(drillbits == null || drillbits.length() == 0){
+ return;
+ }
+ final String[] connectInfo = drillbits.split(",");
+
+ /* For direct connection we can get URL string having drillbit
property as below:
+ drillbit=<ip>:<port> --- Use the IP and port specified as the
Foreman IP and port
+ drillbit=<ip> --- Use the IP specified as the Foreman IP
with default port in config file
+ drillbit=<ip1>:<port1>,<ip2>:<port2>... --- Randomly select the
IP and port pair from the specified
+ list as the Foreman
IP and port.
+
+ Fetch ip address and port information for each drillbit and
populate the list
+ */
+ for(String info : connectInfo){
+ info = info.trim();
+
+ if(info != null){
+ // Split each info to get ip address and port value
+ final String[] drillbitInfo = info.split(":");
+
+ // Check for malformed ip:port string
+ if(drillbitInfo == null || drillbitInfo.length == 0){
+ continue;
+ }
+
+ /* If port is present use that one else use the configured one
+ Assumptions: 1) IP Address provided in connection string is
valid
+ 2) Port without IP address is never specified.
+ */
+ final String port = (drillbitInfo.length == 2) ? drillbitInfo[1] :
config.getString(ExecConstants.INITIAL_USER_PORT);
+ final DrillbitEndpoint endpoint = DrillbitEndpoint.newBuilder()
+ .setAddress(drillbitInfo[0])
+
.setUserPort(Integer.parseInt(port))
+ .build();
+ endpointList.add(endpoint);
+ }
+ }
+ }
+
public synchronized void connect(String connect, Properties props)
throws RpcException {
if (connected) {
return;
}
final DrillbitEndpoint endpoint;
+ final ArrayList<DrillbitEndpoint> endpoints = new ArrayList<>();
if (isDirectConnection) {
- final String[] connectInfo =
props.getProperty("drillbit").split(":");
- final String port =
connectInfo.length==2?connectInfo[1]:config.getString(ExecConstants.INITIAL_USER_PORT);
- endpoint = DrillbitEndpoint.newBuilder()
- .setAddress(connectInfo[0])
- .setUserPort(Integer.parseInt(port))
- .build();
+ // Populate the endpoints list with all the drillbit information
provided in the
+ // connection string
+ populateEndpointsList(endpoints,
props.getProperty("drillbit").trim());
--- End diff --
if "drillbit" is unset in the connection string then this code path won't
be called at all. If "drillbit" string is specified in connection string then
the value is set to empty string when none exists. So .trim() will not cause
NPE. But moved it in callee.
> As per documentation, when issuing a list of drillbits in the connection
> string, we always attempt to connect only to the first one
> -----------------------------------------------------------------------------------------------------------------------------------
>
> Key: DRILL-5015
> URL: https://issues.apache.org/jira/browse/DRILL-5015
> Project: Apache Drill
> Issue Type: Bug
> Components: Client - JDBC
> Affects Versions: 1.8.0, 1.9.0
> Reporter: Sorabh Hamirwasia
> Assignee: Sudheesh Katkam
>
> When trying to connect to a Drill cluster by specifying more than 1 drillbits
> to connect to, we always attempt to connect to only the first drillbit.
> As an example, we tested against a pair of drillbits, but we always connect
> to the first entry in the CSV list by querying for the 'current' drillbit.
> The remaining entries are never attempted.
> [root@pssc-60 agileSqlPerfTests]# /opt/mapr/drill/drill-1.8.0/bin/sqlline -u
> "jdbc:drill:schema=dfs.tmp;drillbit=pssc-61:31010,pssc-62:31010" -f
> whereAmI.q | grep -v logback
> 1/1 select * from sys.drillbits where `current`;
> +-----------------+------------+---------------+------------+----------+
> | hostname | user_port | control_port | data_port | current |
> +-----------------+------------+---------------+------------+----------+
> | pssc-61.qa.lab | 31010 | 31011 | 31012 | true |
> +-----------------+------------+---------------+------------+----------+
> 1 row selected (0.265 seconds)
> Closing: org.apache.drill.jdbc.impl.DrillConnectionImpl
> apache drill 1.8.0
> "a little sql for your nosql"
> This property is meant for use by clients when not wanting to overload the ZK
> for fetching a list of existing Drillbits, but the behaviour doesn't match
> the documentation.
> [Making a Direct Drillbit Connection |
> https://drill.apache.org/docs/using-the-jdbc-driver/#using-the-jdbc-url-format-for-a-direct-drillbit-connection
> ]
> We need to randomly shuffle between this list and If an entry in the shuffled
> list is unreachable, we need to try for the next entry in the list.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
