3

The official document of Elasticsearch says we can use URL as a setting of Read-only repository. The document also says several protocols, including HTTP, is supported for the URL. (refer: https://www.elastic.co/guide/en/elasticsearch/reference/current/modules-snapshots.html#_read_only_url_repository )

I understand that the snapshots are stored in the repository as a directory structure which contains Lucene index and some meta data.

Given above conditions, I think GET request to a directory URL of a HTTP URL repository need to return a list of files/directories inside, in a certain format, for the snapshot API to restore a snapshot. At least, I thought some sort of ls operation on the directory containing Lucene indices were needed.

Questions:

  1. Is my understanding correct?
  2. If so, what is the required format for the directory URL? (What is the required response for HTTP request to the directory which makes snapshot API's restore work?)
Yuki Inoue
  • 197
  • 8

1 Answers1

2

I also asked this question in Japanese SO and got an answer.

The snapshot, when stored, generates some files which stores the names to the data files. Namely: index, metadata-THESNAPSHOTNAME, snapshot-THESNAPSHOTNAME (note snapshot-THESNAPSHOTNAME files are generated in several directories) . From these files the relative paths of the files are generated; thus no need for ls-ing to the directory URL.

So the answer to my questions would be:

  1. No, my assumption was wrong
  2. No need for any response to the directory URL: URLs are constructed by following the specific file's content.
Community
  • 1
Yuki Inoue
  • 197
  • 8