A Load Balanced Service over different sites, howto replicate the docker-volumes?

Hello,
I’m just wondering how you would solve the following problem:

We have 3 sites with two RancherOS Nodes each, in total 6 nodes. We want to implement a redundant, load balanced Web-Service were we need to synchronize/replicate the files and data of the service.
We did some tests with a convoy-glusterfs but it seems just to slow for this.
Is there any possibility to have a master docker-volume and doing some kind of a periodic replication of the data to the other docker-volumes?
How are you solving this problem?

I forgot to say that for us it is important to keep the data on our premise, so we don’t plan to put the data on any cloud-storage except it’s our own storage.

Any hints are very welcome!
Peter

@ApolloDS

Your question raises further questions that I’ve now got to ask. I am in the same boat-- except for the on prem part. We’re hybrid, so I guess I should be more concerned (?)

  1. How slow is too slow?
  2. What alternatives have you looked into?

Please do correct me if I am wrong-- the purpose of glusterfs in your setup is to provide all of the hosts synchronized access to one another’s docker containers, right?

Please don’t hesitate to get ahold of me, e-mail is jake@klouds.org and hangouts is faddat@gmail.com

Good luck, have fun!

Hi Jacob,
Thank you for your answer.
To your question about slow…
Slow is for us when the underlining Filesystem is not answering after 5 seconds. This is happening with GlusterFS but it’s normal since we have spread the Cluster over different sites (testing).
The Webpage is very slow answering then.
I don’t know if GlusterFS has some kind of a mode were you can define a master node and it will then replicate to slave nodes the filesystem content. Now it seems that GlusterFS is always waiting for an answer from all the involved nodes and this is slow when not in the same LAN.

Since we use FreeNAS on every site we were thinking of implementing Convoy-NFS Storage on each site and doing a Master-Slave Replication with Rsync. Until now it is only a plan not tested yet.

What kind of alternatives have you looked into?

Thank you very much!

Peter