Hello
We have restarted our work on updated KNOX 1.4.0 to KNOX 1.6.1 in IBM
Hadoop Execution Engine. Although we are able to run the gateway service
comfortably, we are not able to use the NGINX re-routes that we used to do
earlier.
I will try to explain this below:

HEE: Hadoop Execution Engine (manages remote Hadoop clusters. Supports
running remote jobs, running Spark sessions using LIVY or JEG, remote HDFS
connection, remote HIVE connections, etc.)
CPD: IBM Cloud Pak for Data (IBM Machine Learning Platform that integrates
with other IBM addon services like HEE)
Now, in order to run HEE with CPD, we have to undergo a registration
process that enables a CPD cluster to recognize HEE on a remote Hadoop
Cluster.
As part of this process, we rewrite the KNOX gateway URL using NGINX and
following this, every CURL to the KNOX GATEWAY in HEE is done using the
rewritten NGINX URL.
This is where we see a problem. We also send some headers along with the
CURL request to this rewritten NGINX URL and this used to work fine until
KNOX 1.4.0.
Now, what I suspect is that there has been some change in KNOX 1.6.1 on a
granular level in the way headers are handled and this I say based on:
https://github.com/apache/knox/pull/374/files
This is a critical upgrade for us and any help from your side will be very
much appreciated. I am also available for a quick 15-30 minute call where I
can demo the issue to anyone who can help.

Please let me know if there is anything else needed from my side to
understand the HEE structure or the way KNOX is used in it.

PS. I am not sure if I am subscribed to the mailing list. Could use some
help confirming that.

Regards.
Akshay Kotecha Jain (AK)
Senior Software Developer, IBM Cloud
-- 
Regards
Akshay Kotecha Jain

Reply via email to