Snapshot rollback doesn't work

[Gurnek]

Apache Iceberg version

None

Please describe the bug 🐞

Hello, I have encountered an issue with the rollback functionality. I have my data in a Glue catalog, and I am running the following code in some form:

tbl = catalog.load_table('db.table')
print(tbl.metdata.current_snapshot_id)
tbl.manage_snapshots().rollback_to_snapshot(123).commit() # 123 in this case would be a snapshot id that I know is an ancestor
print(tbl.metadata.current_snapshot_id) # will print the same thing as before

The table will still read from the same snapshot, however if I try to call rollback_to_snapshot(123) again, it will give the error snapshot 123 is not an ancestor of current state: 123.

Any help would be appreciated

Willingness to contribute

• I can contribute a fix for this bug independently
• I would be willing to contribute a fix for this bug with guidance from the Iceberg community
• I cannot contribute a fix for this bug at this time

[geruh]

Hey @Gurnek, thanks for raising.

I tested this against my glue catalog with the 0.11.0 release and haven't been able to reproduce. Could you share some more information to help repo because maybe I'm missing something.

For context I'm running this:

catalog = GlueCatalog('glue', **{'warehouse': 's3://<bucket>/warehouse/'})

schema = Schema(
    NestedField(1, 'id', IntegerType(), required=False),
    NestedField(2, 'data', StringType()),
)

catalog.create_namespace_if_not_exists("db")

tbl = catalog.create_table('db.test_tbl', schema=schema)

for i in range(3):
    tbl.append(pa.table({'id': pa.array([i], type=pa.int32()), 'name': [f'data-{i}']}))

snapshots = tbl.snapshots()
target = snapshots[0].snapshot_id # <-- old snapshot

print(f'Before: {tbl.metadata.current_snapshot_id}')

tbl.manage_snapshots().rollback_to_snapshot(target).commit()

print(f'After:  {tbl.metadata.current_snapshot_id}')

Which then prints:

Before: 6672913416757656702
After:  2295944502128096469

[Gurnek]

Ok, I was investigate the specific error I was getting as well, about the ancestor, and looking at my snapshot list for the table, all of my snapshots have a parent_snapshot_id of None which means they will fail the ancestor lookup. I'm not sure why all of the parents are gone or why I was able to call the rollback function once initially but I'm not convinced its a bug with the library anymore. Thank you