-
Notifications
You must be signed in to change notification settings - Fork 1.2k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
The topic receiver silently failed. isClosed does not reflect #14535
Comments
Hey @bdelville Apologies for the late response here.
Do you mean to say that you are now running with all logs activated? Without client side logs, there is not much we can do. If you have the timestamp on when this happened, you can log an Azure support case where the support folks can then look at service side logs to see if there was anything unexpected that happened during that time on your Service Bus instance. For now we will close this issue. |
I was saying that we are running the instance with debug logs activated for "rhea" and "azure". |
Thanks @bdelville When you get the logs, please open an Azure support ticket with the logs, the timestamp for when you see the issue, your service bus instance details, topic and subscription name. This will allow the support folks to query the service side logs to look for any abnormalities. Also ask them to loop in the Azure SDK team to look at the client side logs that you would be sharing. Feel free to use this issue as a reference. |
Hello, we reproduced and now have the logs. |
Thanks for the logs @bdelville We can start looking into this to look for any client side issues, but I highly recommend opening an Azure support ticket with the details listed in my previous comment #14535 (comment) to rule out service side issues as well as to support you better. |
Thanks for looking into it, |
This PR has a few changes in it, primarily to improve our robustness and our reporting: General reliability improvements: - Migrates to a workflow that treats subscription start as a retryable entity, rather than just link creation (which is what had previously). - It checks and throws exceptions on much more granular conditions, particularly in addCredit - Error checking and handling has been migrated to be in far fewer spots and to be more unconditional, which should hopefully eliminate any areas where an exception or error could occur but it never gets forwarded or seen. SDK debugging: - Adds a new SDK only flag (forwardInternalErrors) which will make it so areas that used to eat errors now can forward them to processError. Prior to this the errors were only logged, but that meant they could be missed. Most of these would probably be considered cosmetic by customers so this is primarly for debugging purposes within the SDK itself. - The internal `processInitialize` handler has been split into two (still internal) handlers - preInitialize and postInitialize. preInitialize runs before init(), and postInitialize runs after init() but before addCredit. This lets us write more reliable tests. These are not exposed to customers. Fixes #14535
Hi @bdelville , we've recently released a fix that should address the error message you saw by being more aggressive about retries: |
Describe the bug
The topic receiver silently failed. Then calling isClosed does not return a disconnected status
To Reproduce
Steps to reproduce the behavior:
Hard to reproduce as it happened without anything special after running for 16 days on Kubernetes.
Actually running with full Azure logs activated to provide extra traces.
Expected behavior
If the receiver get disconnected, either call the MessageHandlers.processError or return true for receiver.isClosed.
This issue was not happening with version 1.x.
The text was updated successfully, but these errors were encountered: