drill-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Taylor Jones <monitor...@gmail.com>
Subject Drillbit hostname configuration error
Date Sun, 19 Oct 2014 20:27:57 GMT
Hi all,

I'm new to Drill, and I'm still trying to get a handle on how all of the
configuration options work. I was able to successfully follow the tutorial
for the embedded setup and I'm trying to get a distributed system set up
now. I thought I should take baby steps, so my environment is very simple:

   - Host A: Zookeeper
   - Host B: Drill
   - Host C: Squirrel SQL + JDBC driver

I'm not using Hadoop yet, since there's only one Drill service running. I
followed all of the instructions, and I was able to successfully connect to
the Drill server using sqlline from Host B. However, when I try to connect
using Squirrel SQL from a remote server (Host C), I get an error.

Using connection URL "jdbc:drill:zk=HostA:2181/drill/drill-cluster" (my
cluster id is "drill-cluster") I get a large stacktrace, the root cause of
which is this:

Caused by: java.util.concurrent.ExecutionException:
java.net.ConnectException: Connection refused: localhost/
at io.netty.util.concurrent.AbstractFuture.get(AbstractFuture.java:37)
... 13 more
Caused by: java.net.ConnectException: Connection refused: localhost/
at sun.nio.ch.SocketChannelImpl.checkConnect(Native Method)
at sun.nio.ch.SocketChannelImpl.finishConnect(SocketChannelImpl.java:739)
... 6 more

Judging from the stacktrace, the JDBC driver is getting told to try to
connect to "localhost" instead of Host B. Thinking this odd, I went into
Zookeeper using zkCli.sh to look at the persisted config. Lo and behold,
"get /drill/drill-cluster/<HostB_znode>" is showing "localhost" as well:

[zk: localhost:2181(CONNECTED) 0] get

localhost���� ��
cZxid = 0x7f
ctime = Sun Oct 19 19:50:28 UTC 2014
mZxid = 0x7f
mtime = Sun Oct 19 19:50:28 UTC 2014
pZxid = 0x7f
cversion = 0
dataVersion = 0
aclVersion = 0
ephemeralOwner = 0x14929cfe9210023
dataLength = 70
numChildren = 0

At this point, I'm fairly sure I this is my problem. It explains why Host B
can connect over sqlline (drillbit is available at localhost) and why I
can't connect from Host C (no drillbit at localhost). I'm still a little
unclear on how the client connection to Drillbits actually works, but I
assume that at some point an external client can connect to it. Judging
from the architecture diagrams, I believe Zookeeper returns the hostname/IP
address of one of the Drillbits to the JDBC driver.

So then, here's my question: How do I go about making the drillbit running
on Host B register itself in Zookeeper with its proper hostname instead of
"localhost"? Please keep in mind I'm new to Drill, so if I've grossly
misunderstood something please enlighten me :)

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message