Documentation needed

[parweb]

It would be cool to have some documentation to test the repo in multi server mode

[brendanodwyer]

Yeah it would be nice to see some more documentation related to how some of the components fit together and how the project is designed to run in multi-server mode. Also any settings or guidelines for running this(non-production) just to try out would be helpful.

[dnr]

Thanks for your interest. More docs would certainly be useful, and I'll try to migrate some of our internal docs in the coming weeks.

It would be helpful to know what sort of environments you're interested in testing things out in. E.g., machines in a rack with ssh access, VMs on EC2, containers in Mesos, etc.?

In any case, you'll probably end up modifying the scripts in scripts for your environment and distributing them with your existing deployment tools. You might need to add a -masters flag (spec) if you can't make the dns name blb-master resolve to something useful.

As far as settings, the defaults should work fine for a test cluster. Most of the tunables can be tuned with config files: use the -masterCfg, -curatorCfg, and -tractserverCfg flags to point to JSON files. The structs are here: master, curator, tractserver. For master and curator, you can tune the raft parameters with a -raftCfg file also.

[brendanodwyer]

Honestly i wanted to test it out on Bare Metal, but also inside of containers. I have seen some presentations and documentation that came from Bloomberg earlier this year about running Hadoop on top of Kubernetes, and i was wondering if perhaps this could run on top of Kubernetes as well.

I was also curious if you guys had broken out list of things that you wanted to finish and build out with this, because the system as a whole seems very interesting and i wouldn't mind helping to take a crack at some of it.

[EasonLiao]

There's a future and contribution section that roughly describes what need to be built, but it's at a very high level and lacking details. If people really want to contribute (personally I would be interested), maybe it's worthwhile to break down them into actionable tickets (issues)?

[brendanodwyer]

Yeah I had noticed the section there at the bottom of the README, but was curious if there was a list of them more broken down and actionable elsewhere.