Open codefromthecrypt opened 4 years ago
Some users want to know how a domain name resolution failed - how many queries were sent and why each of them has failed. The given stack trace shows 3 DNS queries were failed with Network is unreachable
error. Without them, some users would be curious about what the exact cause was.
I agree that the stack trace is too long, though. How could we simplify it without losing much detail?
since you know netty is used. one way is to look for a cause of DnsNameResolverException. then in the WARN level message just add that cause message. In DEBUG one dump everything
because as far as I can tell, there is really no information here needed except the cause which is duplicated multiple times. unless trying to fix the code of the resolver itself, I don't need that stack trace
to put short.. where we own the whole stack, we can maybe be more concise than where 3rd party code is involved. ex stack trace when unknown scenario or where unknown code involved. concise for handled scenario and only known code involved.
warn is fair enough, tells a problem, though ideally doens't affect the already existing resolved ips
anyway that's thoughts so you can decide what is appropriate even if nothing. tx for listening!
One extra context - more volume when responding to a user request is probably not as big of a deal, but I realized this actually continues in the logs a lot (with backoff though) when it's from a DnsAddressEndpointGroup
I see. How about simplifying/omitting the trace when verboseExceptions
is disabled?
is verboseException a thing ?
I think more to point is the message. do you have a reasonable chance to know the problem even if you disable stack traces in log format
On Fri, Aug 2, 2019, 1:48 PM Trustin Lee notifications@github.com wrote:
I see. How about simplifying/omitting the trace when verboseExceptions is disabled?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/line/armeria/issues/1944?email_source=notifications&email_token=AAAPVVY4IQW2CRL3ODVSECLQCPDBBA5CNFSM4IIBXVQ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD3MVPGA#issuecomment-517560216, or mute the thread https://github.com/notifications/unsubscribe-auth/AAAPVVZZU4FHBGM4AOOGQQLQCPDBBANCNFSM4IIBXVQQ .
This WARN will repeat in logs, and since there's no user code involved, and the code that is involved is closely maintained (netty), I'm wondering if it is necessary to get a long stack trace to say unreachable.