Question about how to using noop for deep storage on single node

Background:

  1. We use Druid for performance storage, we hope it both support single node and distribution node.

  2. When using in single node mode, cause of segments(deep storage) and segments-cahce exists, mean local mode(druid.storage.type=local), this will lead our disk doubling.

  3. So when I saw this chapter: ‘http://druid.io/docs/0.12.0/configuration/index.html’ , I’, thinking if there is a way that in single node, we don’t need use segments, but only segments-cache, this can reduce our disk usage.

  4. But when I config like that, I get an error like : ‘Unknown provider[noop] of Key[type=io.druid.segment.loading.DataSegmentPusher, annotation=[none]], known options[[hdfs, s3, local]]’ . here is the detail: https://github.com/druid-io/druid/issues/5687

As reply from gianm, I still have question about it, or may I have misunderstanding about Druid for noop design.

Thanks a lot.~~

As Gian said the noop option is not what you should use, it is there for some dependency injection.
You have 2 options
If you want to have a standalone one node then use local.

If you want a distributed mode then you can use HDFS, S3, GFS but do not use noop it doesn’t get you anywhere

Druid always expects there to be a deep storage, there’s no “local segment cache only” mode, so even if the Noop DataSegmentPusher were usable, your historical nodes would never get any data.