make the Status Provider distributed

[uniray7]

build a Status Provider as a cluster

[uniray7]

the ELK cluster architecture is as below: note: "ES" is elasticsearch

I'll choose 2 kind of ES instance node:

• master node:
  • responsible of cluster-wide operation, like create & delete index(table)
  • reserve some cluster metadata, like shard table
  • route search & indexing and other CRUD request to data node
• data node:
  • responsible of executing those requests routed by master node

I plan building 1 master node & 2 data node on 3 VMs.

@weitingchou: Actually, it's better to add another master node for master failure, but I think we don't need right now. It is for saving cost of aws and I think it may not need to make the cluster too robust. What do u think? @weitingchou: the kibana cost most of memory and cpu resource. I guess master node manage only 2 data nodes, so it may not overburdened to run master node, logstash and kibana on a vm

[uniray7]

build the ES standalone instead of cluster right now. if the scale become larger, then make ES to cluster