2

We are planning to implement hot and cold architecture by allocating shards to different nodes tagged with hot/warm/cold. I want to understand more about performance benefits of this architecture.

Lets say I move last 15 days of my indices to hot nodes,60 days older indices to warm nodes and all other older nodes to cold nodes.

  1. Now If I am executing a search query with a time range of last 15 days, will it search across all nodes or only the nodes tagged with "Hot" nodes?
  2. Shard allocation filter settings applied while writing to a new index are applicable while executing search queries also?
  3. If elastic search is searching across all the nodes to get the result then is there any way I can specify a filter like, "if its last 15 days, get results from only indices/shards present in Hot nodes"?
Sumit Nekar
  • 175
  • 13

1 Answers1

0
  1. The elastic will automatically search all nodes if you specify the index name like "index_name*_search" in the search API. Note it's the * thing which is doing the Job.
  2. Didn't get the question.
  3. See, you just have to provide the "range" parameter in search API, it will get you the data from wherever it is present. So last 15 days data will automatically come from the hot node and 60 days data will come from both hot and warm node. So configure your ILM accordingly.