Azure AI Translator suddenly stopped translating with glossaries
Our translator application was working as expected for multiple months before suddenly after the weekend we started failing translation for cases when glossaries are being passed, with the error being-
'code': 'InvalidGlossaryAccessLevel', 'message': 'Cannot access glossary location with the current permissions.', 'target': 'Operation'
This only applies to the glossary case, as otherwise using the application to translate documents works fine and the SaS permissions we are using for our blobs (including the glossaries) seem to be working as well. The translator is in US West while the blob storage is in US West 3, but considering the source documents are also stored there and are working fine, I don't think that would be an issue.
For further context, we are passing the glossary on a blob level and I have tested with both read=True and read=True, list=True Blob SaS Permissions.
Thanks in advance!
1 answer
-
Anshika Varshney 13,405 Reputation points • Microsoft External Staff • Moderator
Hi Udayan Pandey,
If your Azure AI Translator suddenly stopped translating without any change on your side, this usually points to either a temporary service behavior or a configuration/usage issue, not a complete failure in your code.
What this issue usually means
If translation was working earlier and suddenly stopped, then most common reasons are
- temporary service disruption
- backend update or change in behavior
- authentication or request issue
- known limitation in input format
Azure Translator is continuously updated, and sometimes service behavior can change or be temporarily affected without any change from your side. [github.com]
You can follow these checks step by step to identify the issue quickly:
1. Check service health
First check if there is any outage or service disruption
Azure provides status and service health dashboards to verify this
If there is an ongoing issue, you may see
- temporary failures
- empty responses
- translations not happening
Azure recommends checking service health when behavior changes suddenly [github.com]
2. Test using simple input
Try a very simple request like
Hello world → translate to another language
If this works, then the issue is likely related to your input If it does not work, then it is likely service or configuration related
3. Check request format and parameters
Make sure your request follows the expected API pattern
For text translation API
- use correct endpoint
- include api version
- include target language parameter
Example pattern POST translate API with api-version and to language parameter [docs.azure.cn]
4. Verify authentication
Sometimes translation stops due to auth issues
Check that
- API key is correct
- region header is correct
- no expired or rotated key
Even if the service was working earlier, a key change or config update can break requests
5. Check if input contains mixed languages or unsupported format
Azure Translator has some known limitations
For example
- mixed language sentence may not translate properly
- unusual formatting can give unexpected result
If your input contains multiple languages or special formatting, try simplifying it
Example fix
- split text into single language segments
This is a known behavior in Translator service [learn.microsoft.com]
6. Try from another region or resource
In some cases, issue can be region specific
- test the same request from another resource or region
- compare behavior
If it works elsewhere, then issue is localized
7. Check known issues documentation
Azure maintains a list of known issues and limitations
It is always good to review here before assuming a full outage https://learn.microsoft.com/en-us/azure/ai-services/translator/reference/known-issues
If same request worked earlier, now it intermittently fails or stops translating and there are no changes from your side
then it is often a temporary service-side issue or backend update, which usually resolves automatically after some time
In short
- sudden stop usually means service or backend issue
- check service health first
- verify API key and request format
- test with simple input
- avoid mixed language input
- compare across regions
I hope this helps. If you share your request sample or response error, I can help narrow down.
Thankyou!
-
Anshika Varshney 13,405 Reputation points • Microsoft External Staff • Moderator
Hi Udayan Pandey,
We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution, please do share that same with the community as it can be helpful to others. Otherwise, will respond with more details and we will try to help.
Thankyou!
-
Udayan Pandey 0 Reputation points
Hi, thanks for the help! The problem turned out to be because of glossary names which originally contained spaces. I guess there was a change in validation or request formatting at some point and that broke our existing pipeline. We adjusted the filenames to have _ instead of spaces and that fixed it.
-
Anshika Varshney 13,405 Reputation points • Microsoft External Staff • Moderator
Hi Udayan Pandey,
Thanks a lot for coming back and sharing the update really helpful!
Glad to hear you were able to identify the root cause.
It makes sense that the issue was related to glossary names containing spaces. Azure Translator behavior can sometimes be sensitive to input formatting or validation changes, especially when updates are rolled out on the service side.
Good catch on replacing spaces with underscores. that’s a clean and effective workaround.
This will definitely help others facing similar issues, especially if they are using existing pipelines where naming conventions were not strictly enforced earlier.
Please 'Upvote'(Thumbs-up) and 'Accept' as answer if the reply was helpful. This will be benefitting other community members who face the same issue.
Thankyou!
Sign in to comment