topic Re: Vector Store update stale in Syncing status even the actual sync task is done. in Machine Learning

Vector Store update stale in Syncing status even the actual sync task is done.

airbots — Wed, 19 Aug 2026 18:06:47 GMT

VS status remains "syncing" indefinitely — rechecked after 12 hours, status still "syncing". The Databricks job completes successfully and the new data is queryable:

Re: Vector Store update stale in Syncing status even the actual sync task is done.

Islam_hoti — Thu, 20 Aug 2026 12:39:35 GMT

Hey, @airbots

The index status just mirrors the sync pipeline behind it, so "syncing" forever usually means that pipeline is stuck, not your job. Go to Jobs & Pipelines, find the system-managed pipeline named after your index, and check its event log the actual error is almost always there.

Two things I'd rule out first: make sure delta.enableChangeDataFeed is on for the source table, and check whether your job is doing an overwrite/replace rather than append or merge. Either one breaks the CDF chain and forces a full re-snapshot, which on a big table looks exactly like being stuck.

Also worth running describe() on the index and looking at detailed state instead of the top-level status it's a lot more specific.

Re: Vector Store update stale in Syncing status even the actual sync task is done.

balajij8 — Thu, 20 Aug 2026 17:26:24 GMT

@airbotsThe underlying issue is generally a metadata reconciliation lag where the Vector Search index status remains stuck displaying Syncing in Catalog Explorer even after the sync pipeline has completed. The metadata status reporting mechanism generally runs on a separate path from the actual data-serving and query execution planes. Its generally a cosmetic display glitch rather than a functional error. Since the underlying Databricks job completed successfully with a SUCCEEDED state and your queries are actively returning new data, the index itself is healthy and operational.

To verify your index state without relying on the UI flag, confirm that test queries consistently return your latest payload and verify that the row count matches your source Delta table.

For Continuous sync mode, the backend will continue processing incoming stream updates normally regardless of the visual tag. If the stale status persists across sync cycles, open a ticket providing your Workspace ID and index name for verification. Do not drop or recreate the index as a workaround as it may introduce unnecessary service downtime for an endpoint that is already fully functional and serving traffic.