feat: implement restore operation #1502
Conversation
loleek
requested review from
houqp,
xianwill,
wjones127,
fvaleye,
roeap,
rtyler and
mosyp
as code owners
loleek
changed the title
|
I will file python binding implement in another PR. Is that OK? |
|
I saw integration test failed. But it looks like not caused by my PR. |
| if !ignore_missing_files { | ||
| check_files_available(object_store.as_ref(), &files_to_add).await?; | ||
| } |
There was a problem hiding this comment.
interesting - would not have thought that it is useful, to delivberately create a state with missing files, but it seems the reference implementation also supports this :).
There was a problem hiding this comment.
The data file may missing for many reason, deleted unintentionally or by vacuum operation.This will cause data missing if always allow to restore, so we must make users aware of the problem.
| match try_commit_transaction(object_store.as_ref(), &commit, commit_version).await { | ||
| Ok(_) => {} | ||
| Err(err @ TransactionError::VersionAlreadyExists(_)) => { | ||
| return Err(err.into()); | ||
| } | ||
| Err(err) => { | ||
| object_store.delete(&commit).await?; | ||
| return Err(err.into()); | ||
| } | ||
| } |
There was a problem hiding this comment.
Just making sure I understand. For restore we always want to fail, if any other operation was performed on the table?
There was a problem hiding this comment.
Yes, if any other operation was performed, the restore is invalid because the ADD and REMOVE files are computed based on latest snapshot state and state of restored version.
Indeed it is not, unfortunately we have some flakiness in the tests due to memory issues. |
Description
This is a implementation of the Restore Command.
Related Issue(s)
#837