Skip to content

Conversation

@allenss-amazon
Copy link
Member

@allenss-amazon allenss-amazon commented Oct 9, 2025

fixes #204

@allenss-amazon allenss-amazon changed the title Initial IndexName encode/decode Full support for DB numbers other than zero in cluster mode. Oct 14, 2025
@allenss-amazon
Copy link
Member Author

Fixes issue #204

@allenss-amazon allenss-amazon marked this pull request as ready for review October 14, 2025 00:33
Signed-off-by: Allen Samuels <[email protected]>
Signed-off-by: Allen Samuels <[email protected]>
Signed-off-by: Allen Samuels <[email protected]>
Signed-off-by: Allen Samuels <[email protected]>
Signed-off-by: Allen Samuels <[email protected]>
Signed-off-by: Allen Samuels <[email protected]>
Signed-off-by: Allen Samuels <[email protected]>
Comment on lines 41 to 42
The SchemaManager uses the MetadataManager to manage a single name space of
indexes, i.e., map<string, GlobalMetadataEntry>.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Appreciate the thought going into this. The interoperability is a headache

One question: when someone creates an index in db_num > 0 on Valkey 9, how would a shard running on Valkey 8 behave? I am guessing it would crash, right? I guess that behavior is fine, but we should document this

Another option I guess is that we have namespaces in the metadata manager. Right now we use "vs_index_schema". We could move to "vs_index_schema2"? Nodes on the old versions would just not get the index definitions for those created with vs_index_schema2, which should be fine, and we can register for callbacks of "vs_index_schema" in the new node for backwards compat. Probably a bit more work, but it would prevent the Valkey 8.0 nodes from crashing in such cases

@allenss-amazon
Copy link
Member Author

Revised to reject RDB or cross-shard metadata that contains any object with a semantic version higher than what's understood by the current code.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[BUG] Support multi-db in CME

2 participants