2020-09-25 09:42:44 UTC - Rahul Vashishth: TCP connection to pulsar service url is reachable. i checked <https://github.com/apache/pulsar/blob/master/pulsar-client-tools/src/main/java/org/apache/pulsar/client/cli/PulsarClientTool.java#L116-L119|how >`pulsar-client`<https://github.com/apache/pulsar/blob/master/pulsar-client-tools/src/main/java/org/apache/pulsar/client/cli/PulsarClientTool.java#L116-L119| produce on auth enabled endpoint>. I am also doing the same. ---- 2020-09-25 10:32:02 UTC - Aravindhan: *Pulsar function/source logs in stdout*
<#C5Z4T36F7|general> I am running the pulsar in a docker container. While running the pulsar function and source connector(process/thread runtime) logs are getting placed in the /tmp/functions directory. But couldn't get these in container logs (stdout). Is there any possible way to redirect these logs to docker container logs(stdout)? ---- 2020-09-25 13:20:07 UTC - Guillaume: Hi, I am running pulsar in kubernetes, is it really necessary to use local disk provisionner ? is is better if we use it ? Thanks. ---- 2020-09-25 13:55:04 UTC - Rolf Arne Corneliussen: Is there a way to abort CI test on github that seems to hang forever? ---- 2020-09-25 14:49:19 UTC - Raphael Enns: Hi, I have a couple different pulsar setups running, and I'm seeing high CPU usage from the threads named `bookkeeper-ml-workers-OrderedExecutor-#-0`. I'm running pulsar 2.6.1 in standalone mode. It creates 4 threads with the name above and each thread is constantly at 100%, basically using up 4 out of 8 cores. I stopped all processes that were connecting to pulsar for several minutes, but the CPU usage of pulsar didn't drop at all, but I did notice there were no longer any log lines for those threads while nothing was communicating with pulsar. I'm running pulsar 2.5.2 at another location, but with separate zookeeper, bookkeeper and broker processes. The broker process has 8 of the OrderedExecutor threads, and they are consistently using 20% to 25% CPU each. Do you know what could be causing the high CPU on those threads? Thanks. ---- 2020-09-25 16:22:22 UTC - Jan Wolski: @Jan Wolski has joined the channel ---- 2020-09-25 18:41:59 UTC - Anton Burtsev: @Anton Burtsev has joined the channel ---- 2020-09-25 19:37:27 UTC - Nathan Clayton: @Nathan Clayton has joined the channel ---- 2020-09-25 20:21:08 UTC - Addison Higham: this is a bit tricky to do generally because functions may be forked to a separate process. If you want to open an issue on github there could be a discussion, but you could potentially do it in a custom docker image by having another process `tail` those files and write to `/dev/stdout` ---- 2020-09-25 20:22:47 UTC - Addison Higham: @Guillaume not sure I understand, the helm charts have support for using any storage class, which can allow you to use cloud storage like EBS/GCP persistent disk ---- 2020-09-25 20:28:16 UTC - Addison Higham: That thread pool is used to do most of the work against "managed ledgers" which is the underlying storage for message, so this includes writing messages, rotating ledgers, deleting, etc, etc. It is expected to be somewhat busy, but if it is continuing to be busy minutes after that may mean they are are overhwhelming your bookkeeper tier or saturating your network connecting such that you are getting delays. Are you using batching? That can help reduce work there. ---- 2020-09-25 23:20:42 UTC - Jarrod Johnson: Does anyone have any suggestions / best practices around dead letter and retry topics? I'm curious how people generally handle failures for topics where data loss is a concern. Seems like it would be a huge hassle and maintenance burden to have dead letter and/or retry topics configured for everything topic you cared about. Are there any best practices out there? Any info would be greatly appreciated! ---- 2020-09-26 04:30:28 UTC - N Navin: @N Navin has joined the channel ---- 2020-09-26 07:11:08 UTC - Guillaume: Sorry I wasn't clear. My question is: for a production deployment is it better to use local volume or a Ceph volume is also ok ? By the way is there a guideline for production environement (like RAM, CPU, number of bookies, proxy, broker...) ? ----
