Datastax Enterprise - 5.0 Best Practices - Installation

Question

I am currently evaluating the Datastax Enterprise 5 installation for my production system. There are many methods available for installation. When we choose runinstallter unattended method by DSE using option file it provide two modes 1. Service Based - Need root permission and binaries are installed in /usr/share/dse and /etc/dse. 2. No Service Based - Not need root and binaries can be installed on custom location equivalent to tar based installation without service based.

I have following questions -

1. Is there any best practice available which method is best suited for production installation ( in short any problem in running no service based runinstallter installation)
2. Is there a way we can modify runinstaller in service based installation to point to another dse home then /usr/share/dse and /etc/dse , something like /Cassandra which is owned by casandra user.
3. Any other best practice on the method of installation with is currently live in production without any issues.

Regards

Answer 1

1. Any of the methods specified here are fine for production installations
2. Not that I know of, you might want to look at using the Tarball installation if you need this level of configuration
3. There are a whole lot of things you need to think about when planning a cluster for DSE 5. I would start by looking at this list here.

Answer 2

I'm an OpsCenter developer who works on the Lifecycle Manager feature, so I'm more than a bit biased... but I think that OpsCenter LifeCycle Manager is an excellent way to install and manage DSE if you don't already have something like Chef or Ansible that you use enterprise-wide. It's a graphical webapp with a RESTful API in case you need to do any scripting of it. It deploys DSE using deb/rpm packages over SSH and can configure pretty much every DSE option there is.

As to your other questions:

1. Services vs no-services installations: You probably want a services-based installation. It behaves more like a "normal" linux service that can be managed with the 'service' command. A no-services install is primarily useful if you don't have root access because of very tight security policies in your org, and if you choose to go that route you'll need to decide how you want to manage DSE startup and shutdown (for clean reboots, for example).
2. The DSE installer can probably handle non-standard paths, but I'm not familiar enough with the details. LCM can handle some non-standard paths but not all of them (DSE will always be installed to the standard locations, for example). If you want to very tightly control every aspect of the install, tarball is your best choice. That's a lot of complexity, though, do you REALLY need to control every path?
3. The OpsCenter Best Practice service is probably the best list of recommended things to do in Prod, and is very easy to turn for LCM-managed clusters. But even if you don't use LCM, I recommend you set up OpsCenter so you can use the Best Practice Service.

You can find the OpsCenter install stesp at: https://docs.datastax.com/en/latest-opsc/opsc/LCM/opscLCMOverview.html.