ES 1.7 data loss


(Gunho Shin) #1

Hello, I using ES 1.7 for log storage.
I have 1 master(no data) & 1 data nodes.
My app has been separated for write and search.
Today, I found something strange.
Data has been loss depend on writing position.

case 1

  • Writing app send bulk request to master node (Java API).
  • Searching app send search request to master node (REST API).
    = data loss occurred (few hundred per minute).

case 1

  • Writing app send bulk request to data node (Java API).
  • Searching app send search request to master node (REST API).
    = data loss not occurred.

I using default config for my cluster.
I'm confused this situation. somebody help?


(David Pilato) #2

I'm pretty sure you have something in logs.

That being said, why this architecture?

I'd go either with one node only (no redundancy no HA) or 3 nodes (2 data and one master only).


(system) #3