FWIW, this problem was resolved by using the IP address for the S3 target 
instead of the hostname.  No idea why hostname wouldn't work here--the hostname 
is completely valid and resolves forward and backwards from the drill machine.  
Nitish, the author of the article, was kind enough to help me on the Minio 
slack channel.  He is not quite sure why Drill is behaving this way but thinks 
perhaps the fs.s3a.path.style.access flag in core-site.xml only works when 
using an IP address.  Either I missed something when reading the Drill docs or 
possibly this needs to be explained somewhere.

Thanks,
Clark

On 2/13/19, 5:18 PM, "Updike, Clark" <[email protected]> wrote:

    I'm trying to run Drill 1.15 against an S3 compatible Minio instance 
following the steps described here:
    https://blog.minio.io/query-minio-datastore-with-apache-drill-dcaf71d0cee5
    
    So on my Minio server, minio1, I have a bucket, drillbucket1.  I'm putting 
in core-site.xml that my fs.s3a.endpoint is minio1:9000.  In my S3 storage 
plugin, named s3-drillbucket1, I'm setting "connection": "s3a://drillbucket1".  
That is, unless I missed something, I'm set up just like the article.  But when 
I go to run a query, I get this error:
    
    Error: SYSTEM ERROR: UnknownHostException: drillbucket1.minio1: Name or 
service not known
    
    Can someone explain what I have misconfigured that is causing the bucket to 
become part of the server name?  Or perhaps the setup has changed since the 
article was written?
    
    Thanks,
    Clark
    

Reply via email to