Improve documentation on snapshot restore using URL (read-only) repository

[ppf2]

A few documentation improvement suggestions based on user experience in the field:

• Will be nice to add a cross reference to setting up read only url repository from the restore section of the guide.
• Providing an example (eg. for file protocol) will be nice.

PUT /_snapshot/my_restore_repo
{
  "type": "url",
  "settings": {
    "url": "file:///nfs/path"
  }
}

• When setting up a read-only repository, only setting up repositories.url.allowed_urls is sufficient. The statements below suggests that path.repo is also required.

"URL Repository supports the following protocols: "http", "https", "ftp", "file" and "jar". URL repositories with http:, https:, and ftp: URLs has to be whitelisted by specifying allowed URLs in the repositories.url.allowed_urls setting. "
"URL repositories with file: URLs can only point to locations registered in the path.repo setting similar to shared file system repository."

[frederikbosch]

I dont get the purpose of this method at all. Could someone dxplain it?

[clintongormley]

@frederikbosch you can make snapshots available to the public (or privately to others) via a read-only URL repository.

[frederikbosch]

@clintongormley Alright, but how is that achieved? Is the data pushed to some other location?

[frederikbosch]

My case is that I want to push the backup directly (without hitting the local fs) to another location (over HTTPS over FTPS/SFTP). And I thought that this backup type might be useful. For me S3, HDFS and Azure are not an option.

[clintongormley]

@frederikbosch no, it's read only. it's for making read only copies available to others, not for making the snapshots yourself. we don't have an http only writable repository

[frederikbosch]

@clintongormley Thanks for your answers! That is really helpful.

[BaranOrnarli]

@clintongormley is there no way to transfer a local zipped folder (generated via _snapshot "fs") to ES Master node, then distribute it across the ES cluster without "shared drives"? I was looking for a way to simply point to a "local filesystem repo" and have the Master ES node distribute it into the cluster or _snapshot-restore it by reading the local filesystem (read-only). The "Type: 'fs'" ES-repo requires a COPY of the folder in EACH datanode/masternode which is really difficult with one folder/zip that is gigantic.

[debadair]

[docs issue triage]