Files
scylladb/idl
Yaniv Michael Kaul 3cba27d25f topology: propagate error messages through raft_topology_cmd_result
When a topology command (e.g., rebuild) fails on a target node, the
exception message was being swallowed at multiple levels:

1. raft_topology_cmd_handler caught exceptions and returned a bare
   fail status with no error details.
2. exec_direct_command_helper saw the fail status and threw a generic
   "failed status returned from {id}" message.
3. The rebuilding handler caught that and stored a hardcoded
   "streaming failed" message.

This meant users only saw "rebuild failed: streaming failed" instead
of the actionable error from the safety check (e.g., "it is unsafe
to use source_dc=dc2 to rebuild keyspace=...").

Fix by:
- Adding an error_message field to raft_topology_cmd_result (with
  [[version 2026.2]] for wire compatibility).
- Populating error_message with the exception text in the handler's
  catch blocks.
- Including error_message in the exception thrown by
  exec_direct_command_helper.
- Passing the actual error through to rtbuilder.done() instead of
  the hardcoded "streaming failed".

A follow-up test is in https://github.com/scylladb/scylladb/pull/29363

Fixes: SCYLLADB-1404

Closes scylladb/scylladb#29362
2026-05-11 17:01:15 +03:00
..
2026-03-18 19:24:26 +01:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00
2026-04-12 19:46:33 +03:00