[00:16:57] 10AI-Governance, 10Lift-Wing, 10ORES, 10artificial-intelligence, 10Machine-Learning-Team (Active Tasks): Create Draft Model Deployment Guidelines - https://phabricator.wikimedia.org/T276598 (10ACraze) @Chtnnh - some answers for your questions > Will these discussions be asynchronous? Yes, the community... [06:06:14] (03PS1) 10Ladsgroup: Disable precaching for all wikis except enwiki [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/672881 [06:51:11] (03CR) 10Elukey: "Amir can you add a little more info, for context, about what is the expected result after this change?" [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/672881 (owner: 10Ladsgroup) [06:56:28] 10Lift-Wing, 10Machine-Learning-Team, 10Patch-For-Review: Create ml-serve k8s cluster - https://phabricator.wikimedia.org/T272918 (10elukey) @ACraze thanks for the explanation, it seems that we have a good motivation to setup Istio with service mesh support, medium to long term it will be needed so it makes... [08:49:23] 10Lift-Wing, 10Machine-Learning-Team, 10Patch-For-Review: Create ml-serve k8s cluster - https://phabricator.wikimedia.org/T272918 (10Joe) FWIW, ServiceOps decided against using a full mesh networking for our services because we considered istio to be both very complex and not really needed for our level of c... [11:14:13] 10Lift-Wing, 10Machine-Learning-Team, 10Patch-For-Review: Create ml-serve k8s cluster - https://phabricator.wikimedia.org/T272918 (10elukey) >>! In T272918#6920560, @Joe wrote: > FWIW, ServiceOps decided against using a full mesh networking for our services because we considered istio to be both very complex... [17:40:15] elukey: thanks for the chat, i agree with skipping the full service mesh install for our proof of concept and then revisiting later down the road [17:40:49] Is a full service mesh install a requirement for an MVP? [17:41:14] i don't think so [17:41:51] kfserving docs say we can skip it too: https://github.com/kubeflow/kfserving#prerequisites [17:42:15] Okay cool, the focus is getting the MVP done, and I'm do want the emphasize the "M" in MVP. If something is a requirement for "M" then we need to do it, but if not, we definitely should not do it. [17:42:15] chrisalbon: my doubt was if even the basic use case, like a model (micro-service) trying to contact a pre/post processor would have needed (from the KFserving pov) a service mesh [17:42:46] pre-post processing is definitely a requirement for "M" [17:44:00] exactly, but after a chat with accraze it seems that we'd be good without istio configured in a full mesh, that should ease a little bit the work that we have to do after k8s [17:44:14] so knative + istio (no service mesh) + KFServing [17:44:21] (that is already a lot) [17:44:27] I don't mean to rehash the conversation you both had. Pre/Post processing is a requirement for the MVP, so as long as that is possible then won't micromanage [17:45:18] yeah if necessary we should be able to manually inject sidecars (or maybe even use envoy which other teams at WMF already use & support) [17:45:37] cool, cool [17:52:24] chrisalbon: nono I didn't mean that, I like to add my thoughts in IRC so others can chime in :) [17:53:10] No that is wonderful and advances the foundation's goal of transparency, I just wanted to make sure we'd have pre/post processing in the MVP. [17:54:03] today I also found https://www.oreilly.com/library/view/kubeflow-operations-guide/9781492053262/ [17:54:22] nice! [17:54:44] that seems something for poor SREs trying to figure out how it works (and loosing faith in everything in the meantime) [17:55:04] "Install Kubeflow on an existing on-premise Kubernetes cluster" [17:55:15] lol, yeah I want to read that too [17:57:06] I'd love to have David Aronchick talk to the team too as part of the speaker series. [17:57:36] He cofounded Kubeflow [17:58:52] would love to attend that talk [18:06:05] yeah that would be really great [18:12:56] wow [18:14:22] chrisalbon: I'd also love to have an intro about Kubeflow from Holden Karau! (maybe connected to other things like Spark, feature stores, etc..) [18:20:35] +1