Thanks Mohamed for the updates!
I really do not want to rain on your parade, but I am still not sure
whether moving those methods from the StreamsMetricsImpl to the
StreamsMetrics is the right approach to get rid of
ProcessorContextUtils#getMetricsImpl().
I do also not agree about the
Thanks Mohamed,
I think Bruno raised a good point in the ticket that the
node name is not well known from within the Processor at the
time of init(), so users would basically have to make up a
name to pass into the sensor. Maybe this is ok, but it
doesn't seem too nice.
Offhand, it seems like
KIP updated with the comments of Bruno Cardona.
Le 06/07/2020 à 22:36, Mohamed Chebbi a écrit :
Thank Bruno for your review.
Changes was added as you sugested.
Le 06/07/2020 à 14:57, Bruno Cadonna a écrit :
Hi Mohamed,
Thank you for the KIP.
Comments regarding the KIP wiki:
1. In section
Thank Bruno for your review.
Changes was added as you sugested.
Le 06/07/2020 à 14:57, Bruno Cadonna a écrit :
Hi Mohamed,
Thank you for the KIP.
Comments regarding the KIP wiki:
1. In section "Public Interface", you should state what you want to
change in interface StreamsMetrics. In your
Hi Mohamed,
Thank you for the KIP.
Comments regarding the KIP wiki:
1. In section "Public Interface", you should state what you want to
change in interface StreamsMetrics. In your case, you want to add two
methods. You can find a good example how to describe this in KIP-444
https://cwiki.apache.org/confluence/display/KAFKA/KIP-639%3A+Move+nodeLevelSensor+and+storeLevelSensor+methods+from+StreamsMetricsImpl+to+StreamsMetrics