Azure APIM response code 0

Ferenc Szabó 66

We are seeing requests in API Management logged in Application Insights with the response code:
0 [not sent in full (see exception telemetries)]
On the client side, the response is "503 Service Unavailable". We are not really sure what this phenomenon is, what's in between the client and the APIM we are not aware of that's canceling the request and responding with the 503.
Any insights into this?

ChaitanyaNaykodi-MSFT 23,106 Reputation points Microsoft Employee

2021-04-28T18:16:42.107+00:00

Hello @Ferenc Szabó , apologies for the delay. AFAIK response code 0 might have multiple reasons behind it as it basically means as "undefined". If possible can you please run a trace for that particular call? as it will help in pin pointing the exact issue. Please let me know if there are any concerns. Thank you!
Ferenc Szabó 66 Reputation points

2021-05-01T16:47:16.517+00:00

Hello Chaitanya,

Sorry for the late response, we were very busy with our launch. What we are seeing in App Insights are requests received by the Api Management and forwarded to the backend service which results with an exception called "ClientConnectionFailure at forward-request"

But we also see traces like

And we are not really sure whether these are actually cancelled by the end users (as some stack overflow threads would indicate) or a network error on our side.
Would exception handling and/or retry logics mitigate these?

Thank you,
Feri
Ferenc Szabó 66 Reputation points

2021-05-12T16:02:56.777+00:00

Hi @ChaitanyaNaykodi-MSFT , any ideas? Interestingly enough we are seeing this much more often from our Samsung Tizen and the LG webOS clients than the SPA clients, which indicates this is a client side issue, even though it seems like the APIM loses connection to the backend App Service.

Thank you,
Feri
ChaitanyaNaykodi-MSFT 23,106 Reputation points Microsoft Employee

2021-05-13T00:20:48.107+00:00

Hello @Ferenc Szabó , my sincere apologies for the delay. The observation made is correct. ClientConnectIonFailure means client connected to APIM aborted pending request. In the case APIM aborts corresponding request to backend. From the screenshots above we can see that the duration is not that high for the failed responses as well, so it does look like Client side issue as observed by you. I think reproducing the error from client side might help in determining the root cause of the issue. Please let me know if there are any concerns.
Ferenc Szabó 66 Reputation points

2021-05-13T06:26:38.427+00:00

Hi Chaitanya,

What's confusing is that in the first case the backend doesn't seem to be called at all, that's why we thought that there's a connectivity issue between the APIM and the backend. But I guess the best solution would be checking from the client-side.
But aside from this, we would be interested in our options of troubleshooting API errors. We don't see more details in the logs when we change the Verbosity to Verbose from Information. Are we missing some logs? We also have the "Diagnostic Settings" enabled, but the GatewayLogs does not contain anything useful either.

Thank you!
Feri

1 answer

ChaitanyaNaykodi-MSFT 23,106 Reputation points Microsoft Employee

2021-05-13T22:49:38.407+00:00

Hello @Ferenc Szabó , this is the data which gets added to Application insights. You can also add trace policy to do custom logging for the API. Additionally in the advanced options while setting up application insights you can decide on how backend & frontend requests/responses are logged and please be aware of performance implications of logging in APIM.

Also regarding the verbosity level if this is the case please be aware that the setting applied at individual API level will over-ride the setting for All API levels.

You can also refer to this tutorial to understand more about monitoring published API's. Please let me know if there are any concerns. Thank you!
Please sign in to rate this answer.

0 comments No comments
Sign in to comment