2018-11-23 09:16:04 UTC - Meilong Huang: @Meilong Huang has joined the channel ---- 2018-11-23 09:42:04 UTC - bossbaby: Hi all, i don't understand cluster and node in pulsar, why 1 cluster something 3 node to work. Please, Someone explain help me ---- 2018-11-23 09:42:26 UTC - Nguyễn Hữu Phong: @Nguyễn Hữu Phong has joined the channel ---- 2018-11-23 09:44:05 UTC - Zhang Pengfei: @Zhang Pengfei has joined the channel ---- 2018-11-23 09:57:19 UTC - jia zhai: @bossbaby Here is a brief architecture: <http://pulsar.apache.org/docs/en/concepts-architecture-overview/> ---- 2018-11-23 09:57:49 UTC - jia zhai: you may find some concept in this page. ---- 2018-11-23 10:15:03 UTC - bossbaby: @jia zhai broker is node, it right? ---- 2018-11-23 10:15:13 UTC - jia zhai: yes ---- 2018-11-23 10:20:06 UTC - bossbaby: i think create more broker will increase performance ---- 2018-11-23 15:56:21 UTC - Harald Gustafsson: Hi all, I just got the python pulsar client to consistently seg fault when subscribing to a topic. This happens from one process, always the same. While another process (a test) manage to subscribe to the same topic. Both have unique subscriber id. Is there any (simple) way of enabling more logging inside the pulsar client. A similar tests works when run locally instead of in the k8s cluster. The Pulsar log is have an exception logged: ---- 2018-11-23 15:56:47 UTC - Harald Gustafsson: ---- 2018-11-23 16:00:17 UTC - Harald Gustafsson: I can also see that the container is restarted after a while. ---- 2018-11-23 16:00:32 UTC - Harald Gustafsson: The deployment is made with the helm chart ---- 2018-11-23 16:46:59 UTC - Harald Gustafsson: seems a log4cxx config file can be passed in, anyone know if the python wheel (on mac or linux) is compiled with the USE_LOG4CXX flag on? And does anyone have a working config? The repo have one, I changed from INFO to DEBUG but no change in the log output. ---- 2018-11-23 16:55:06 UTC - Matteo Merli: @Harald Gustafsson the Python wheels are not compiled with log4cxx . Is the segfault happening when the corresponding “authentication required” error is triggered by broker? ---- 2018-11-23 16:56:52 UTC - Harald Gustafsson: I can't be sure that they happen at the same time, they are running in different containers. But it is at least happening during the time I run the test, I don't have any other users active. ---- 2018-11-23 16:57:27 UTC - Matteo Merli: any way to get a core dump? ---- 2018-11-23 16:59:30 UTC - Harald Gustafsson: I'm not sure, I have tried finding a method to get a core dump from a k8s container but not found anything yet. Since this only happens when run in the cluster it is a bit harder to debug. ---- 2018-11-23 17:05:14 UTC - Matteo Merli: Unfortunately, the simple logger only prints put info logs ---- 2018-11-23 17:06:42 UTC - Matteo Merli: and while in c++ client lib it’s possible to get a callback when logs are getting printed, that’s not exposed in Python binding ---- 2018-11-23 17:13:50 UTC - Harald Gustafsson: OK, I think most users would be happy with just being able to set the log level for the python client. For example when things work I would prefer to use log level ERROR, but for debugging setting DEBUG. ---- 2018-11-23 17:17:22 UTC - Matteo Merli: Yes, that would be better. We’ll add that ----
