Understanding metrics wrt eg dispatcher/stream and others


I am experiencing ChannelClosedExceptions and therefore aim to do a retry when this happens if it is caused by a connection level failure that the client wishes not to see any consequences of.

When I look at admin/metrics.json I see a couple of similar yet different lines containing ChannelClosedExcpetion and I would like to understand how they relate.

I am extra interested in the difference between



(See below for the list of different metrics that I want to understand)

The reason I am particularly interested in the difference between the two above is that when I try out a ResponseClassifier using TimeoutException a get ChannelClosedException logged upon timeout and
in this case it’s only the /dispatcher/stream type of metric:

rt/http-out/client/#/io.l5d.fs/asia:4141/dispatcher/stream/failures/com.twitter.finagle.ChannelClosedException 3

In this case there is not really a connection level channel closing, it’s only the Linkerd timeout that is shorter than the respone time of the service. (The ChannelClosedException get logged at debug level)

rt/https-out/service/svc/www.asiakastieto.fi:443/failures/com.twitter.finagle.ChannelClosedException 1
rt/https-out/service/svc/www.asiakastieto.fi:443/sourcedfailures/#/io.l5d.fs/www.asiakastieto.fi:443/com.twitter.finagle.ChannelClosedException 1
rt/https-out/service/svc/www.registret.dk/failures/com.twitter.finagle.service.ResponseClassificationSyntheticException 2
rt/https-out/client/#/io.l5d.fs/www.asiakastieto.fi:443/service/svc/www.asiakastieto.fi:443/failures/com.twitter.finagle.ChannelClosedException 1
rt/https-out/client/#/io.l5d.fs/www.asiakastieto.fi:443/failures/com.twitter.finagle.ChannelClosedException 1

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.