You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
From the last few days of testing with Hono I think we need some kind of "Hono error codes". Which then translate into HTTP error codes. Yet, those Hono specific error codes should be reported back in e.g. the HTTP response (of possible by the underlying protocol).
Hono
HTTP
Text
Actual problem
1001
503
no credit available for sending request
The RequestResponse client has no credit to send a request to the tenant service
2001
503
no credit available for sending request
The RequestResponse client has no credit to send a request to the registry service
3001
503
no credit available for sending request
The RequestResponse client has no credit to send a request to the credentials service
1002
503
request timed out after 200ms
A request to the tenant service timed out after 200ms
2002
503
request timed out after 200ms
A request to the registry service timed out after 200ms
3002
503
request timed out after 200ms
A request to the credentials service timed out after 200ms
Of course this table is not complete, but only there to show the problem.
In the result you will always get "503" and a text which isn't helping either. Sometimes you run into "temporarily unavailable", which again masks out a complete set of problems.
Yes, there is the console/log. But if you have a sporadic problem and 150 pods of the HTTP protocol adapter running, then the logs are useless. As is OpenTracing/Jaeger, as you can't really catch all requests and store them forever.
So I think it would make sense to define some kind of Hono specific error. That error may by itself contain a mapping to a HTTP error code, or at least this could be done in a single location, in order to be consistent. For HTTP the error code could be reported back using an HTTP property in the response header.
The text was updated successfully, but these errors were encountered:
From the last few days of testing with Hono I think we need some kind of "Hono error codes". Which then translate into HTTP error codes. Yet, those Hono specific error codes should be reported back in e.g. the HTTP response (of possible by the underlying protocol).
Of course this table is not complete, but only there to show the problem.
In the result you will always get "503" and a text which isn't helping either. Sometimes you run into "temporarily unavailable", which again masks out a complete set of problems.
Yes, there is the console/log. But if you have a sporadic problem and 150 pods of the HTTP protocol adapter running, then the logs are useless. As is OpenTracing/Jaeger, as you can't really catch all requests and store them forever.
So I think it would make sense to define some kind of Hono specific error. That error may by itself contain a mapping to a HTTP error code, or at least this could be done in a single location, in order to be consistent. For HTTP the error code could be reported back using an HTTP property in the response header.
The text was updated successfully, but these errors were encountered: