Support Questions

pollard · ‎09-14-2018

Hi,

I have just successfully run a Docker container image of Cloudera. I can connect to the instance using Safari on the Mac. However, I get an error message that says it can't connect:

"Could not connect to quickstart.cloudera:21050 (code THRIFTTRANSPORT): TTransportException('Could not connect to quickstart.cloudera:21050',)"

Also, when I do a check configuration it comes up with a bunch of errors:

hadoop.hdfs_clusters.default.webhdfs_url	Current value: http://localhost:50070/webhdfs/v1 Failed to access filesystem root
OOZIE_EMAIL_SERVER	Email notifications is disabled for Workflows and Jobs as SMTP server is localhost.
Hive	Failed to access Hive warehouse: /user/hive/warehouse
HBase Browser	Failed to authenticate to HBase Thrift Server, check authentication configurations.
Impala	No available Impalad to send queries to.
Hadoop Security: Sentry Service	Failed to connect to Sentry API (version 1).
Hadoop Security: Sentry Service	Failed to connect to Sentry API (version 2).

Anyone have a direction they can point me on how to resolve all of these issues. I thought the Docker image would work out of the box.

Thanks!

bgooley · ‎09-18-2018

@pollard,

Cool, I worked with Apple for a bit in 2005ish too. 🙂

To business, looks like it was Docker memory default to 2GB that was the cause. I forgot that I noticed my container was sluggish so I bumped my Docker memory to 4GB. I just prooved it by reverting to 2GB and then to 4GB as explained below:

Here is my session:

(1)

$ docker images
REPOSITORY TAG IMAGE ID CREATED SIZE
<none> <none> 6d2b3c820712 19 hours ago 7GB

(2)

docker run --hostname=quickstart.cloudera --privileged=true -t -i -p 8888:8888 -p 80:80 -p 7180:7180 6d2b3c820712 /usr/bin/docker-quickstart

(3)

(in container shell)

# sudo -u hdfs hdfs dfs -ls /
18/09/18 16:57:09 WARN ipc.Client: Failed to connect to server: quickstart.cloudera/172.17.0.2:8020: try once and fail.
java.net.ConnectException: Connection refused
at sun.nio.ch.SocketChannelImpl.checkConnect(Native Method)
at sun.nio.ch.SocketChannelImpl.finishConnect(SocketChannelImpl.java:739)

(4)

I log into Hue as cloudera:cloudera

Check Config shows all the failures you mentioned in your first post

(5)

start over

$ exit (in the terminal launched)

* basically, kill the running container

(6)

In Docker UI configuration, Advanced:

Set Memory to 4.0GiB

Click the Apply & Restart button

(7)

Run:

$ docker run --hostname=quickstart.cloudera --privileged=true -t -i -p 8888:8888 -p 80:80 -p 7180:7180 6d2b3c820712 /usr/bin/docker-quickstart

(8)

now everything works:

# sudo -u hdfs hdfs dfs -ls /
Found 6 items
drwxrwxrwx   - hdfs supergroup          0 2017-10-23 23:46 /benchmarks
drwxr-xr-x   - hbase supergroup          0 2018-09-18 17:06 /hbase
drwxr-xr-x   - solr solr                0 2017-10-23 23:48 /solr
drwxrwxrwt   - hdfs supergroup          0 2018-09-18 17:06 /tmp
drwxr-xr-x   - hdfs supergroup          0 2017-10-23 23:48 /user
drwxr-xr-x   - hdfs supergroup          0 2017-10-23 23:48 /var

(9)

Hue is fine... can run queries

NOTE: the 80 and 7180 ports aren't really necessary for exposure to host OS for this test case.

I hope this solves the issue for you too!

If so, it is pretty lame that lower memory would somehow let servers start and listen but then cause issues for networking to those listening ports. I guess that is a welcome to docker for me 🙂

View solution in original post

pollard · ‎09-18-2018

Yeah, they sold Powerschool off to a competitor who took it way past what we had. One of the Java engineers talked the new company into letting him re-write it in Java.

It looks pretty cool from the last time I looked at it.

I remember that group back in the day. I still have my email address and am using 50GB of iCloud storage for the day my Mac goes down. I bought a 2016 laptop in 2017 and absolutely love it. All except the memory. In which they fixed in the newest releases of MacBook Pro. Dang company! 🙂

Dude, I got some stories of sticking with something. I have about 33 years of Unix experience starting back in the day with Sun Microsystems servers. I own a Sun Sparc Classic and a Sun Enterprise 250 workgroup server. Both of them will still boot.

I wil let you know if anything pops up as I am going to do some Django development as a phase II project to manage users automatically. I figured out how to web scrape the Hue page using the older user list URL (/useradmin/users/) that doesn't use Javascript to load the user list (/hue/useradmin/users/). Although, Hue warns the older URL is old, it still works. I have not figured out yet how to load the user list with the new (current) URL of "/hue/useradmin/users/" but will get back to it later. It would be nice to have an API for Hue as Cloudera Manager has. My project was to auto-delete Cloudera and Hue users that have left the company. I have all the prototype functionality built and am about to bring it all together into a package.

Thanks for all your help!

Cloudera Community

Support Questions

Docker Cloudera image Hue login issues