Why wait between master - client/data?

[zetaab]

https://github.com/pires/kubernetes-elasticsearch-cluster/blame/master/README.md#L56-L60

Could you explain why you have written this part? I have now tried 5 hours to get this setup crashing without waiting, and still it works. I have also deployed client and data nodes (and wait that those are running) before master and after that created master nodes - still it works. The thing in elasticsearch data/client is that it will automatically try to reconnect to masters if it does not have that connection. It means that when client/data nodes are started without master - it will join to masters after those are available. Of course data insert/queries does not work before masters are online. This is just my experience, I tried my best to break it. I propose that you could remove that wait thing after master.

Why I propose this: people does not understand that this elasticsearch cluster installation is actually automatical (https://www.reddit.com/r/kubernetes/comments/63ockr/anyone_using_rediselasticsearchkafka_on/) "example 1 has some stars on repo, seems to work, but then again, involves manual steps, eg: Wait until es-master deployment is provisioned"

at least you could mention that "if you have problems wait between master and client/data, but it should work without that as well"

[pires]

Interesting. But I never said (wrote) that pods would crash if at least one master wasn't present, right? 🤔

I find it somewhat of a paradox that people want to run single-step deployments in production environments. It's like stating "I don't know nor care what it does, I just want it to run and work for my production scenarios.". But then it doesn't work and they get pissed off at you. Been there, done that, got the t-shirt.

I think that user in particular and many others should look at operators which are much smarter than yaml based recipes.

[pires]

But thank you for raising this. I'll leave it open just in case someone wants to contribute something more.