Affects: Spring Framework Web 6.1.8
I use the ReactorNettyClientRequestFactory within RestTemplate and RestClient.
RestTemplate and RestClient take care of mapping IOExceptions to ResourceAccessExceptions.
In case of using the reactor.netty.http.client.HttpClient, ReactorNettyClientRequest, ReactorNettyClientResponse -
RuntimeExceptions like IllegalStateException (caused by .block() calls), io.netty.handler.timeout.ReadTimeoutException etc. bubble thru - up to RestTemplate and RestClient consumers which should only know the RessoureAccessExceptions.
So most likely IOExceptions should be thrown in ReactorNettyClientRequest, ReactorNettyClientResponse methods.
Furthermore the behaviour is not consistent with other ClientHttpRequestFactoryimplementations
like JettyClientHttpRequestFactory, OkClientHttpRequestFactory, ...
Comment From: jhoeller
A revision will be available in the upcoming 6.1.9 snapshot, consistently converting to IOExceptions. Feel free to give it an early try!
Comment From: helmut-hackl-dynatrace
big thanks! i will give it a try ...
Comment From: helmut-hackl-dynatrace
I patched your changes for testing - looks ok. thanks again.
Comment From: helmut-hackl-dynatrace
one update:
in your fix
you throw in ReactorNettyClientRequest::convertException - after falling thru - the IOException this way:
return new IOException(ex.getMessage(), cause);
In case of eg. a io.netty.handler.timeout.ReadTimeoutException both are null - the message plus the cause.
So you loose the connex to ex and that it was a timeout that caused the exception.
Consumers like org.springframework.web.client.DefaultRestClient.DefaultRequestBodyUriSpec#createResourceAccessException
don't have this information at hand and throw a ResourceAccessException with "null" message.
Comment From: jhoeller
Good point! Revised for 6.1.11 in #33080, including the original exception if there is no cause now.