HPE UNIFIED ANALYTICS SETUP PROBLEM

[Mirza12332]

While installation of HPE Unified Analytics software platform, we got errors. Logs are attached.

[support_s]

System recommended content:

1. Troubleshooting Guide for HPE ProLiant Gen10 and Gen10 Plus Servers

Please click on "Thumbs Up/Kudo" icon to give a "Kudo".

Thank you for being a HPE valuable community member.

[Mirza12332]

We could not find necessary page about above errors. Can you forward to us ?

[sturns]

Hi Mizra,

It took me a couple of goes initially for UA until i discovered I was working off the wrong installation doco  (face palm).

Are you using 1.2 build and installation guide https://docs.ezmeral.hpe.com/unified-analytics/12/index.html ?

Another thing I found 'interesting' was the requirement, as recommended by HPE support to me and also the docs, was to run the installation as the root user.  For me this involved allowing root ssh access to my cluster nodes for the installation.

[Mirza12332]

Hello,

I followed documentation which you sent ( https://docs.ezmeral.hpe.com/unified-analytics/12/index.html )

We installed Rhel 8.8 and all machines have last OS versıon (yum update completed)

Because we did not have kubernetes and machines install kubernetes itselfs with script,  we setup all machines while took reference  as "Table 3. Mount Point Requirements for Kubernetes Hosts in documentatıon" (https://docs.ezmeral.hpe.com/unified-analytics/12/Installation/prereq-install-pph.html).

or should we took reference " Table 2. Mount Point Requirements for Controller and Shadow Controller Hosts" ?

Also we updated .yaml file which defined cluster ip's, username and password to root password, username and ip's.

or Should we reach machine as root and continue installation steps as root ?

Are those steps true ? if there is a mistake about our installation steps please response me and tell us true installation steps please.

Thank you.

[sturns]

I think you are confusing the individual host requirements with the collective host requirements.

With UA we have a control plane and a worker plane.  The control plane has the coordinator node and the controller node.  The worker plane has the k8 workers.  Table 2 lists the build partition requirement for your controller plane and table 3 lists the build requirement for the worker plane. 

The Launcher host (installer node GUI) mentioned in table 1 is the host that you run the installer gui on which connects to the other hosts as root via ssh pub/priv key or static credentials.  For the ssh as root to work you need to configure sshd_config on your hosts to accept root authentication.  Perhaps test you can ssh around your nodes via root prior to you build.  

As long as you enable ssh on your control plane and worker plane you don't need to touch those hosts at all during the installation phase.  All the configuration is done via the Launcher host.  

[Mirza12332]

As you said, we set the cp nodes according to table 2 and the worker nodes according to table 3, and also adjusted the sshd_config settings. We have successfully tested that we can ssh through root. But we still get the same error.

[sturns]

Seems like it's still stuck initially trying to communicate to your nodes.  If you have confirmed that your ssh as root works from your installer node to your control plane and worker plane and you've entered your ssh creds per the doco for the build during the installer process then I would suggest checking the basics such as access to the installation repositories from the installer node (I guess just internet access generally), partitioning requirement nailed down across the cluster, and DNS for the host resolution also.  

If you haven't already, it was recommended by HPE techs to me to destroy the current build on failure and redo the installation.  So for me that wasn't to much of an issue being a new build managed via IaC with Terraform.  So with that in mind I would recommend destroying your current build completly and starting fresh.