Enhance `locate-object` output #1713

New issue

Open

opened 2025-04-10 00:41:56 +00:00 by elebedeva · 0 comments

elebedeva commented

2025-04-10 00:41:56 +00:00

Member

Related to #1635

Currently, the frostfs-cli control locate-object command returns a flat list of shards containing an object, but lacks important details when the object is either:

Split into multiple parts
Stored under erasure coding (EC) policy

In these cases, users cannot determine:

Which specific shard contains which part of the object
The mapping between object parts/chunks and their respective storage locations

Suggested output example:

Object: <object_id>
  Parts:
    - <part_id_1> / <chunk_id_1>
      Shards:
        - shard_id_1
        - shard_id_2
    - <part_id_2> / <chunk_id_2>
      Shards:
        - shard_id_3

Describe the solution you'd like

I see two possible solutions:

1. On `frostfs-cli` level

Modify ListShardsForObject() to return ECInfoError and SplitInfoError
In CLI command:
- Make standard shard location request
- If errors indicate split/EC object:
  - Extract parts/chunks information from errors
  - Make additional HEAD requests for each part (like in object nodes)
  - Aggregate and display complete mapping

2. On `Control service` level

Enhance ListShardsForObject() to:

Return structured response with complete mapping (object/part/chunk to the list of shards containing it)
- In Storage Engine iterate through node's shards
- If we get ECInfoError and/or SplitInfoError, break iterations
- Extract psrts/chunks from errors
- Iterate through list of shards for every found part/chunk
Pretty print response in CLI

Describe alternatives you've considered

Leave everything as is.

## Is your feature request related to a problem? Please describe. Related to #1635 Currently, the `frostfs-cli control locate-object` command returns a **flat** list of shards containing an object, but lacks important details when the object is either: 1. Split into multiple parts 2. Stored under erasure coding (EC) policy In these cases, users cannot determine: - Which specific shard contains which part of the object - The mapping between object parts/chunks and their respective storage locations Suggested output example: ``` Object: <object_id> Parts: - <part_id_1> / <chunk_id_1> Shards: - shard_id_1 - shard_id_2 - <part_id_2> / <chunk_id_2> Shards: - shard_id_3 ```  ## Describe the solution you'd like I see two possible solutions: ### 1. On `frostfs-cli` level 1. Modify `ListShardsForObject()` to return `ECInfoError` and `SplitInfoError` 2. In CLI command: - Make standard shard location request - If errors indicate split/EC object: - Extract parts/chunks information from errors - Make additional `HEAD` requests for each part (like in `object nodes`) - Aggregate and display complete mapping ### 2. On `Control service` level Enhance `ListShardsForObject()` to: - Return structured response with complete mapping (object/part/chunk to the list of shards containing it) - In `Storage Engine` iterate through node's shards - If we get `ECInfoError` and/or `SplitInfoError`, break iterations - Extract psrts/chunks from errors - Iterate through list of shards for every found part/chunk - Pretty print response in CLI  ## Describe alternatives you've considered  Leave everything as is.