2020-09-08 01:51:35 UTC - Dominic Kim: Yes Naver is running a public cloud service with OpenWhisk internally. https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599529895022900?thread_ts=1599491430.016500&cid=C3TPCAQG1 ---- 2020-09-08 04:13:37 UTC - Rajitha Warusavitarana: I tried with a general adobe account. Got this error in the console. `main.min.js:1 unrecognized domain. Please email <mailto:marketingt...@adobe.com|marketingt...@adobe.com>.` https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599538417023100?thread_ts=1599491490.017500&cid=C3TPCAQG1 ---- 2020-09-08 04:30:57 UTC - Rajitha Warusavitarana: Since Apache OpenWhisk builds its components using containers, is it possible to connect GPU or TPU enabled containers and Kubernetes clusters for the Open Whisk project. Such as <https://docs.nvidia.com/deeplearning/frameworks/user-guide/index.html|Nvidia docker>, <https://cloud.google.com/tpu/docs/kubernetes-engine-setup|GKE with TPUs> and <https://cloud.google.com/kubernetes-engine/docs/how-to/gpus?hl=zh-tw|GKE with GPUs>. I understand that this might need a certain amount of work to implement GPU support in a serverless platform. But is it something possible, since I currently don't have much knowledge about the architecture of the OpenWhisk platform I thought of asking this question from the community. Also can we adjust the time-out of a serverless action (serverless function). My aim is to enhance this platform to run long running serverless functions with GPU and TPU support which will benefit scientific applications. https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599539457028900 ---- 2020-09-08 12:06:05 UTC - Michele Sciabarra: I personally like a lot the idea of using OpenWhisk with GPU but keep in mind that the serverless model limits the time an action can run and this may conflict with the requirements of GPU based applications like Deep Learning +1 : Rajitha Warusavitarana https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599566765030600 ---- 2020-09-08 13:33:41 UTC - Rajitha Warusavitarana: Thanks for your feedback @Michele Sciabarra. Sure, I'll keep that in mind. Btw is it possible to configure a time-out when I'm going to enhance the platform, or is there a specific maximum limit like in AWS lambda functions where it cannot be ever changed ? https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599572021034200?thread_ts=1599572021.034200&cid=C3TPCAQG1 ---- 2020-09-08 13:36:16 UTC - Michele Sciabarra: I believe the right approch would be using a “provider”, something rquesting a computation and getting ther results as a trigger +1 : Rajitha Warusavitarana https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599572176035600 ---- 2020-09-08 13:36:35 UTC - Michele Sciabarra: something similar to what kafka or the alarm packages does https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599572195036000 ---- 2020-09-08 13:37:02 UTC - Michele Sciabarra: it is an exciting challenge but no one so far undertake it https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599572222036500 ---- 2020-09-08 13:37:49 UTC - Keerthi Kumar S R: There are two limits • one at action level • another at platform level which is the maximum limit I guess based in the use case support maximum limit can be varied +1 : Rajitha Warusavitarana https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599572269036600?thread_ts=1599572021.034200&cid=C3TPCAQG1 ---- 2020-09-08 13:57:51 UTC - Rajitha Warusavitarana: Thank you @Michele Sciabarra @Keerthi Kumar S R, I'll look into this more and let know about the progress :+1: https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599573471038100 ---- 2020-09-08 21:05:50 UTC - Tyson Norris: Something in #4958 has broken builds with test failures in ContainerProxyTests, I’m trying to fix it. +1 : Rodric Rabbah https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599599150038800 ---- 2020-09-08 22:12:44 UTC - Brendan Doyle: Hi I believe this commit from June introduced a non-backwards compatible change for live clusters. If you try to upgrade the controllers or invokers past this commit, the controller can't talk to the non upgraded invoker or vice versa because of the new json message types introduced here. They will fail to deserialize.
<https://github.com/apache/openwhisk/commit/8759cadeea6101a3693aac1d34d4069f516d5558> https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599603164041400?thread_ts=1599603164.041400&cid=C3TPCAQG1 ---- 2020-09-08 23:00:14 UTC - Rodric Rabbah: this is true https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599606014041500?thread_ts=1599603164.041400&cid=C3TPCAQG1 ---- 2020-09-08 23:00:51 UTC - Rodric Rabbah: i think this is can be fixed by making the deserializer tolerate both old and new formats https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599606051041700?thread_ts=1599603164.041400&cid=C3TPCAQG1 ---- 2020-09-08 23:02:02 UTC - Brendan Doyle: yea I'm looking at the code trying to come up with something. I think we can make a simple fix that will require ordering, but can still safely upgrade. i.e. you need to upgrade the invokers first and then the controllers. https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599606122041900?thread_ts=1599603164.041400&cid=C3TPCAQG1 ---- 2020-09-08 23:41:34 UTC - Brendan Doyle: here's a fix that requires upgrading invokers then controllers. Explanation is in pr description <https://github.com/apache/openwhisk/pull/4970> https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1599608494042100?thread_ts=1599603164.041400&cid=C3TPCAQG1 ----
