You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
fix: Remote signing with table specific s3.signer.endpoint (#2005)
<!--
Thanks for opening a pull request!
-->
<!-- In the case this PR will resolve an issue, please replace
${GITHUB_ISSUE_ID} below with the actual Github issue id. -->
<!-- Closes #${GITHUB_ISSUE_ID} -->
# Rationale for this change
Currently pyiceberg remote singing only works if the sign endpoint is
shared by all tables in a REST Catalog.
However, some Catalogs use table specific endpoints.
If table specific endpoints are used, Pyiceberg sends the sign request
for the second table that is queried to the sign endpoint of the first
table.
The reason for this is, that although we [re-register() a new signer
with different
properties](https://github.com/apache/iceberg-python/blob/996a7ba4dbf4afdb3d46689f1715206b1c355f2a/pyiceberg/io/fsspec.py#L166),
this statement has no effect the second time it runs because of the
unique_id, even if the properties are different.
https://github.com/boto/botocore/blob/8c517320c6a40cd91e8e7fbb05e27183ba2f6dce/botocore/hooks.py#L310-L312
This PR first unregisters the old handler, before adding the new one.
# Are these changes tested?
No. Any idea how we could test them?
Just tested against LAKEKEEPER in a notebook while implementing table
specific endpoints.
# Are there any user-facing changes?
It works now!
<!-- In the case of user-facing changes, please add the changelog label.
-->
0 commit comments