Upstream retries

[andrew-tevent]

I'm occasionally seeing these errors in my SignalR logs for the upstream calls

Failed invoking '[url-redacted]' for 'hub:[hub-name-redacted],event:disconnected,type:sys,category:connections'.","type":"MessagingLogs","collection":"Serverless"

I am at a loss as to why - because I am not seeing any other downtime on the destination URLs or any failures in that destination server etc.. (Any suggestions as to why this is failing?)

But it got me thinking that the Upstream should really have retries on failure?

I know that the EventGrid integration stuff has retries, so I could change to move that instead, but then it would mean that local Emulator would no longer work in my application if I stop using upstream.

[zackliu]

Retry makes sense when the request is totally failed. Like socket failure and we don't get any http response. That can be put on our roadmap.

[andrew-tevent]

I think standard practice is failed responses, and specific Http Status Codes as per the Polly Transient Error retry policy - it includes HttpResponseException and then the specific status codes.