2019-09-03 11:30:29 UTC - bsideup: Hi! Any known memory leaks in the client version 2.4.0? ---- 2019-09-03 12:59:44 UTC - Retardust: Hm, documentation page is not working <https://pulsar.apache.org/admin-rest-api/?version=2.4.0> :disappointed: +1 : Poule ---- 2019-09-03 13:40:08 UTC - Sijie Guo: There is a recent change that broke the css file location. There is a PR fixing it. We are about to merge it soon. +1 : Bruno Bonnin, Poule, Retardust ---- 2019-09-03 13:42:18 UTC - Sijie Guo: I know @Penghui Li was looking into high memory usage in broker side. But I haven’t heard about client side issues. ---- 2019-09-03 13:52:22 UTC - bsideup: Ok, thanks! We’re investigating a memory leak and trying to narrow the scope. Will report if we find anything in the client :+1: ---- 2019-09-03 13:53:02 UTC - Alexandre Nardelli Escobosa: @Alexandre Nardelli Escobosa has joined the channel ---- 2019-09-03 13:59:33 UTC - Sijie Guo: awesome. feel free to ping here if you need any help ---- 2019-09-03 14:01:27 UTC - Penghui Li: @bsideup Whether you are using avro schema? A known problem is related to the reader cache. ---- 2019-09-03 14:01:46 UTC - bsideup: no, no schema is used ---- 2019-09-03 14:03:37 UTC - Penghui Li: Ok, if find anything, ping here please. Thanks for the feedback ---- 2019-09-03 14:07:44 UTC - bsideup: Will do! ---- 2019-09-03 18:00:08 UTC - Rajiv Abraham: Hi, I had a question regarding pulsar functions. Since it has some inspiration from AWS Lambda, AWS Lambda though stateless, keeps the function instance in memory for a while after executions until it finds that it is no longer in use and then ejects it out from memory. This allows us to use resources like database connections and keep the state for more than one invocation if the function is repeatedly called. Note, AWS Lambda gives no guarantees as to how long they keep the function in memory but it's a useful thing in practice. Just to give an e.g. ```python db = Connection() def handler(event, context): # use db ``` re: Pulsar Functions, can I expect similar behaviour or perhaps better? :slightly_smiling_face: ---- 2019-09-03 18:06:54 UTC - Poule: @Rajiv Abraham i need to test this also i hope it works ---- 2019-09-03 18:16:23 UTC - David Kjerrumgaard: @Rajiv Abraham If you deploy your Pulsar functions in K8s, then they each run as their own pod and will continue to run until you explicitly scale the number of instances down or delete the function altogether. +1 : Ming Fang ---- 2019-09-03 19:26:51 UTC - Poule: @David Kjerrumgaard and not k8s what happens? ---- 2019-09-03 19:27:12 UTC - Poule: I do not want the extra k8s movinf ---- 2019-09-03 19:27:16 UTC - Poule: moving part ---- 2019-09-03 19:28:24 UTC - David Kjerrumgaard: If you deploy them as processes on the function worker nodes or as processes on the Broker nodes, then the same is true. They are independent processes that will continue to run unless and until they are stopped. ---- 2019-09-03 19:30:33 UTC - Jon Bock: Pulsar Functions workers are long-running, so state that is local to a function worker will be accessible (to the same worker, not to other workers) until the next restart of the function worker. ---- 2019-09-03 19:35:09 UTC - Rajiv Abraham: thanks @Jon Bock :slightly_smiling_face: ---- 2019-09-03 19:38:54 UTC - Rajiv Abraham: @David Kjerrumgaard Thanks for that info. I'll be deploying on K8. Is it possible for it to be a python process per function instead of a pod per function? I ask because we'll have many pulsar functions, often very simple transformations and a pod with an entire os in it per pulsar function may be a bit heavy in terms of resources ---- 2019-09-03 19:40:47 UTC - David Kjerrumgaard: @Rajiv Abraham Yes, but you will have to configure function worker pods to host the function instances. <https://pulsar.apache.org/docs/en/functions-worker/#docsNav> ---- 2019-09-03 19:41:03 UTC - David Kjerrumgaard: which is a complex process in K8s. :smiley: ---- 2019-09-03 19:44:02 UTC - Rajiv Abraham: haha. complex means job security(or insecurity? :slightly_smiling_face: ). Thanks for the link. I'll read it up. ---- 2019-09-03 19:46:27 UTC - David Kjerrumgaard: complex means networking complexity at the K8s layer :smiley: ---- 2019-09-03 19:48:20 UTC - Rajiv Abraham: haha. my worst nightmare but coffee helps coffee : David Kjerrumgaard ---- 2019-09-03 21:17:14 UTC - Rajiv Abraham: I just realized that my comments above may be construed against Pulsar which was not the intention at all. It was more the intricacies of devops. ---- 2019-09-04 03:24:54 UTC - lionleo: @lionleo has joined the channel ---- 2019-09-04 03:36:38 UTC - David Kjerrumgaard: @Rajiv Abraham No worries..... :smiley: ---- 2019-09-04 07:44:29 UTC - dba: Hi guys. In regards to the many new Txn-commands, should the ProtocolVersion have a v15? ---- 2019-09-04 07:48:30 UTC - Sijie Guo: yes +1 : dba ----
