How to approach dramatically degraded performance in Azure Vector Index

Question

How to approach dramatically degraded performance in Azure Vector Index

Hessel Wellema 276

Starting approximately two hours ago, the average latency of search queries against one of my indexes has jumped from ±1 second to well over two minutes. The index contains ±500 000 documents, all comfortably within the limits of my current plan. I query the index using the Node.js SDK:

"@azure/search-documents": "^12.0.0-beta.2"

Nothing has changed in my codebase. The only recent update was the addition of roughly 100 new documents about six hours ago—several hours before the slowdown began.

What is the best way to diagnose the root cause of this sudden performance degradation? Any guidance is appreciated.

Siva Nair 345 Reputation points Microsoft External Staff Moderator

2025-07-30T10:38:54.8466667+00:00

Hi Hessel Wellema,

Just checking back to see if the above response was helpful, if not let me know any concern you have, happy to help you further.

Thanks!
Hessel Wellema 276 Reputation points

2025-07-30T11:00:30.1+00:00

Hi Siva,

tx. I contacted Microsoft Support and they were able to pinpoint the root cause.
I deleted and added a couple of hundreds of documents hours before the degradation occurred.
Apparently these actions have a negative effect on the performance even hours after the event.

I added an additional replica which should hel address this.

Thank you
Siva Nair 345 Reputation points Microsoft External Staff Moderator

2025-07-31T09:00:38.5566667+00:00

Hi Hessel Wellema,

Hope you got your issue resolved , glad that you got the rootcause and will make it by adding replica as you mentioned, thanks for sharing , this will be really helpful for others who has similar issue.

Best regards!

Accepted answer

0 additional answers

Your answer

Siva Nair 345 Reputation points Microsoft External Staff Moderator

2025-07-30T10:38:54.8466667+00:00

Hi Hessel Wellema,

Just checking back to see if the above response was helpful, if not let me know any concern you have, happy to help you further.

Thanks!
Hessel Wellema 276 Reputation points

2025-07-30T11:00:30.1+00:00

Hi Siva,

tx. I contacted Microsoft Support and they were able to pinpoint the root cause.
I deleted and added a couple of hundreds of documents hours before the degradation occurred.
Apparently these actions have a negative effect on the performance even hours after the event.

I added an additional replica which should hel address this.

Thank you
Siva Nair 345 Reputation points Microsoft External Staff Moderator

2025-07-31T09:00:38.5566667+00:00

Hi Hessel Wellema,

Hope you got your issue resolved , glad that you got the rootcause and will make it by adding replica as you mentioned, thanks for sharing , this will be really helpful for others who has similar issue.

Best regards!

Answer 1

Hi Hessel Wellema,

Rootcause by hessel wellema-

I deleted and added a couple of hundreds of documents hours before the degradation occurred. Apparently these actions have a negative effect on the performance even hours after the event. I added an additional replica which should hel address this.

other workaround -

First lets check the Azure Service Health dashboard to ensure there are no active incidents or service disruptions in your region that might be impacting performance. Next, review your Search service metrics in the Azure Portal under Monitoring. Key metrics to look at include Search Latency, Throttled Queries, Indexing Latency, and Query Volume. These insights can help determine whether the issue is related to backend throttling or a spike in load.

Additionally, try running a simplified version of your query using Search Explorer in the Azure Portal. For example, test a basic query like search=* with a reduced $top value such as 5. This helps isolate whether the latency is caused by specific filters, scoring profiles, or result sizes. You should also review your current query structure in code- if your query uses large $top values, full-text search on large fields, or complex filters, these can dramatically increase query time, especially as your index grows.

check the index statistics to see if there has been a sudden spike in document size or index fragmentation. This can be done via the Azure Portal or through the Index Stats API. Given that you’re using a beta version of the SDK (^12.0.0-beta.2), I would recommend upgrading to the latest stable version to rule out any potential issues or bugs in the beta release. You can do this by running:

npm install @azure/search-documents@latest

If the problem persists after these steps, enable diagnostic logging via Azure Monitor to track query performance and identify problematic requests. Logs can be filtered using Log Analytics to show long-running or failed queries.

Let me know if you need further assistance.

Share via

How to approach dramatically degraded performance in Azure Vector Index

0 additional answers

Your answer