Open TTrapper opened 3 months ago
Thanks! will look into this asap
had the same issue on being stuck at 50 for backfill.
updateTrigger and updateTask work fine tho, so current workaround for me is to just update the documents in the backfill to trigger it
Any updates on this? I'm still getting the same behavior.
I'm posting an update because the behaviour looks a little different now:
1) The _firestore-vector-search index still shows RUNNING with only 50 docs processed, and the rest queued for BACKFILL.
2) When I run the ext-firestore-vector-search-queryCallable
function I get results from more than just the first 50 documents
3) Many more documents now have a vector
field (not just the first 50), but there are still many that don't have it. There doesn't seem to be an ordering to this: documents without the vector field are randomly dispersed among documents that that don't have a vector field.
[REQUIRED] Step 2: Describe your configuration
[REQUIRED] Step 3: Describe the problem
The extension installs and starts to backfill, but it gets stuck after 50 items and just hangs there forever. I've been back and forth with google cloud support for a couple of weeks and they sent me here. Here is a screenshot from the firestore UI:
I've also commented on this seemingly related thread: https://github.com/GoogleCloudPlatform/firebase-extensions/issues/360#issuecomment-2170458164
I don't see an index created under Vertex AI vector search (I assume that's where it would go).
The logs for the funtion
ext-firestore-vector-search-backfillTask
don't show any error:Steps to reproduce:
Expected result
Vectors are computed for all entries in the database
Actual result
Vectors are computed for the first 50 entries and then it hangs.