Snapshot restore requests allow you to restore data stream alias with conflicting name

[jrodewig]

Elasticsearch version (bin/elasticsearch --version):
Version: 8.1.0-SNAPSHOT, Build: default/tar/e38cadb5283feeea06f09d555dbe47fb1d5e8b39/2021-11-22T22:35:22.825882Z, JVM: 17.0.1

OS version (uname -a if on a Unix-like system):
Darwin 20.6.0 Darwin Kernel Version 20.6.0: Mon Aug 30 06:12:21 PDT 2021; root:xnu-7195.141.6~3/RELEASE_X86_64 x86_64

Description of the problem including expected versus actual behavior:

I expect ES to reject snapshot restore requests that restore a data stream alias with the same name as an existing data stream or index.

However, ES currently accepts these requests. This cluster now has a data stream/index and alias with the same name. A user may not be sure which resource a search request targeting the name will hit.

Steps to reproduce:

1. Register a snapshot repository

PUT /_snapshot/my_repository
{
  "type": "fs",
  "settings": {
    "location": "my_backup_location"
  }
}

2. Create a data stream.

POST logs-my_app-default/_doc
{
  "@timestamp": "2099-05-06T16:21:15.000Z"
}

3. Add a logs alias to the data stream.

POST _aliases
{
  "actions": [
    {
      "add": {
        "index": "logs-my_app-default",
        "alias": "logs"
      }
    }
  ]
}

4. Create a snapshot containing the data stream and its alias

PUT _snapshot/my_repository/my_snapshot?wait_for_completion=true

5. Delete the data stream. This also deletes the alias.

DELETE _data_stream/logs-my_app-default

6. Create another data stream with same name as the deleted alias (logs).

PUT _index_template/my-index-template
{
  "index_patterns": ["logs"],
  "data_stream": { },
  "priority": 500
}

POST logs/_doc
{
  "@timestamp": "2099-05-07T16:21:15.000Z"
}

7. Restore the data stream and logs alias from the snapshot.

POST /_snapshot/my_repository/my_snapshot/_restore?wait_for_completion=true
{
  "indices": "logs-my_app-default",
  "include_aliases": true
}

I expected this to return an error due to a conflict between the restored logs
alias and the pre-existing logs data stream. However, the request was
accepted, and both resources now exist.

Retrieve the logs data stream:

GET _data_stream/logs

Retrieve the logs alias:

GET _alias/logs

A search on logs only hits the logs data stream:

GET logs/_search

However, a search on log* hits both the alias and the data stream:

GET log*/_search

[elasticmachine]

Pinging @elastic/es-data-management (Team:Data Management)

[elasticmachine]

Pinging @elastic/es-distributed (Team:Distributed)

[martijnvg]

(deleted comment was meant to be posed to #80976)

[martijnvg]

Actually I think my comment makes more sense here than in #80976.

[jrodewig]

Thanks again @martijnvg. However, I don't think the issue here is with merging aliases. It's that a restored data stream alias can have the same name as an existing data stream or index. I've updated the description to better reflect that.

This same issue doesn't seem to occur for index aliases. For example:

1. Register a snapshot repository

PUT /_snapshot/my_repository
{
  "type": "fs",
  "settings": {
    "location": "my_backup_location"
  }
}

2. Create index1.

POST index1/_doc
{
  "@timestamp": "2099-05-06T16:21:15.000Z"
}

3. Add the logs alias to index1.

POST _aliases
{
  "actions": [
    {
      "add": {
        "index": "index1",
        "alias": "logs"
      }
    }
  ]
}

4. Create a snapshot containing index1 and its alias.

PUT _snapshot/my_repository/my_snapshot?wait_for_completion=true

5. Delete the index1. This also deletes the logs alias.

DELETE index1

6. Create a new index named logs.

POST logs/_doc
{
  "@timestamp": "2099-05-07T16:21:15.000Z"
}

7. Restore index1 and the logs alias from the snapshot.

POST /_snapshot/my_repository/my_snapshot/_restore?wait_for_completion=true
{
  "indices": "index1",
  "include_aliases": true
}

This results in an error:

{
  "error" : {
    "root_cause" : [
      {
        "type" : "illegal_state_exception",
        "reason" : "index, alias, and data stream names need to be unique, but the following duplicates were found [logs (alias of [index1/3IRS4FwJRaOZautW1DW1dA]) conflicts with index]"
      }
    ],
    "type" : "illegal_state_exception",
    "reason" : "index, alias, and data stream names need to be unique, but the following duplicates were found [logs (alias of [index1/3IRS4FwJRaOZautW1DW1dA]) conflicts with index]"
  },
  "status" : 500
}

I'd expect a similar error when restoring a data stream alias with the same name as an existing index or data stream.

[martijnvg]

Thanks for @jrodewig for explaining this to me. This is a bug and needs to be fixed.