I am sure it's not the network problem. Because some changes can successfully
send to other nodes.
So, I think it's the program sometimes can not reponse status with 200,
perhaps 500. But there are no more specific error message to verify that. So
I suggest if you can add retry logic and narrow the try catch scope to log
more detailed message
View this message in context:
Sent from the Apache Kylin mailing list archive at Nabble.com.