[BUG] fix persistent HNSW parameter migration

[codetheweb]

Description of changes

Now correctly handles migration of persistent HNSW params.

Test plan

I updated the cross_version_persist test to parameterize across with_persistent_hnsw_params = [False, True]. Before the changes to chromadb/db/mixins/sysdb.py, the updated test was failing.

[github-actions]

Reviewer Checklist

Please leverage this checklist to ensure your code review is thorough before approving

Testing, Bugs, Errors, Logs, Documentation

• Can you think of any use case in which the code does not behave as intended? Have they been tested?
• Can you think of any inputs or external events that could break the code? Is user input validated and safe? Have they been tested?
• If appropriate, are there adequate property based tests?
• If appropriate, are there adequate unit tests?
• Should any logging, debugging, tracing information be added or removed?
• Are error messages user-friendly?
• Have all documentation changes needed been made?
• Have all non-obvious changes been commented?

System Compatibility

• Are there any potential impacts on other parts of the system or backward compatibility?
• Does this change intersect with any items on our roadmap, and if so, is there a plan for fitting them together?

Quality

• Is this code of a unexpectedly high quality (Readability, Modularity, Intuitiveness)

[HammadB]

If someone upgraded to 0.5.4, and the migration already ran, will it have to be rerun?

[codetheweb]

If someone upgraded to 0.5.4, and the migration already ran, will it have to be rerun?

yes, @atroyn do you know where would be a good place to hook in for this?

[atroyn]

If someone upgraded to 0.5.4, and the migration already ran, will it have to be rerun?

yes, @atroyn do you know where would be a good place to hook in for this?

I think this will have to re-run if it's already migrated, yes. The reason is the migration (which calls _insert_config_from_legacy_params on read) only checks for a None collection config string. Users who have already migrated will have this populated, so the migration will not trigger. This is my fault for missing that the persistent HNSW params were extracted on a different codepath.

The fix is going to be to make sure that the HNSWConfiguration represented in the CollectionConfiguration string contains all the HNSW metadata keys. This is annoying because it means we have to check metadata on every read.

I am not sure we want to eat that performance hit for the possibly 0 users who will have ever changed hnsw:batch_size or hnsw:sync_threshold since these are completely undocumented parameters, for which we otherwise set defaults.

[codetheweb]

Reasoning through it (lost some context from when I worked on this last week):

Without this fix, chromadb will completely refuse to fetch a collection with persistent parameters:

Users who utilized the hnsw config parameters and upgraded to 0.5.4 would have been blocked by this, so I don't think we need to worry about a case where the migration was run before this bug was fixed.

I wonder if it'd be worth adding telemetry for config fields so we have more data in the future.

[atroyn]

I'm unclear on why this is triggering since the migration is nondestructive?