| *** sameo has quit IRC | 00:32 | |
| kata-irc-bot | <caoruidong> One way is editing toml file in /opt/kata/share/defaults/kata-containers | 07:54 |
|---|---|---|
| *** sameo has joined #kata-general | 08:45 | |
| *** sgarzare has joined #kata-general | 08:54 | |
| *** gwhaley has joined #kata-general | 09:08 | |
| kata-irc-bot | <graham.whaley> Razva - as Frank says, if you want to set the defaults globally for all kata containers, edit the global config file. If you want to do it on a per-container basis then you do it the same as docker default - with the docker command line -m and -cpu type options. | 09:50 |
| *** gwhaley has quit IRC | 11:47 | |
| *** gwhaley has joined #kata-general | 13:04 | |
| *** devimc has joined #kata-general | 14:40 | |
| *** lpetrut has joined #kata-general | 15:12 | |
| Razva | is there a Slack channel we can use? Because I can see that IRC is kinda ... empty... :) | 15:38 |
| kata-irc-bot | <salvador.fuentes> Razva: yeah, here the link: http://bit.ly/katacontainersslack | 15:41 |
| gwhaley | there is slack indeed Razva - link is on https://katacontainers.io/community/ irc is normally a bit busier than this though, but we have folks in many timezones, so it comes and goes | 15:42 |
| gwhaley | :snap: | 15:42 |
| gwhaley | but, at least slack waits for folks to log back in and holds your answer for you for a bit ;-) | 15:42 |
| Razva | :) ok. So should I use IRC or Slack? For me it's the same... :) | 15:42 |
| gwhaley | Slack probably has more benefits. We have a bot that copies between the two - so, either works... | 15:43 |
| kata-irc-bot | <razvan.rosca> Yeah, it seems I already had an account here :P | 15:44 |
| *** Razva has left #kata-general | 15:44 | |
| kata-irc-bot | <razvan.rosca> Hi again :slightly_smiling_face: | 15:44 |
| gwhaley | tada! | 15:45 |
| kata-irc-bot | <razvan.rosca> Here's an interesting thought/idea I had during this weekend. As a startup "cloud provider" wannabee, why shouldn't I use Kata instead of KVM? 1) beacause you can't do Windows. Ok, I can live with that. Those who need Windows will get a "full-blown" VM. But...is there a reason #2? | 15:46 |
| kata-irc-bot | <razvan.rosca> I - as a wannabe cloud provider - have a single "concern": 1) isolate clients as best as possible; 2) sell as many VMs per bare-metal as possible. | 15:46 |
| kata-irc-bot | <razvan.rosca> Does Kata fit the bill, or are there any things I'm not getting/realizing? | 15:47 |
| kata-irc-bot | <razvan.rosca> ...anybody....? :slightly_smiling_face: | 15:52 |
| kata-irc-bot | <lpetrut> I think it depends on the type of service you want to provide. Is the customer supposed to use container/orchestrator (e.g. k8s) APIs or just have direct access to the vms? | 15:52 |
| kata-irc-bot | <razvan.rosca> @lpetrut most need/ask for direct access. | 15:52 |
| kata-irc-bot | <razvan.rosca> Aka "Tarzan gives OS, Jane gives money". | 15:53 |
| kata-irc-bot | <razvan.rosca> But what's the diff between a full-blown VM with CentOS Minimal installed vs a Docker CT that uses a minimal CentOS? | 15:54 |
| kata-irc-bot | <razvan.rosca> both are basically using KVM | 15:54 |
| kata-irc-bot | <lpetrut> kata's a container runtime that uses lightweight vms for isolation purposes, not sure how that helps if you need to expose the vms directly. you can have similar lightweight vms without actually using kata. | 15:54 |
| kata-irc-bot | <razvan.rosca> Well ... CloudStack is not something very "lightweight" nor very easy to manage. Networking in Docker is way more "manageable" (for example). Also OS install - even via cloudinit - is a pain compared with a docker hub pull | 15:56 |
| kata-irc-bot | <lpetrut> those are different type of services. fwiw, you may also consider openstack | 15:57 |
| kata-irc-bot | <lpetrut> there's a bit of a learning curve, it may not be worth the hassle for really small environments but it's great nonetheless | 15:59 |
| kata-irc-bot | <razvan.rosca> I've made an OpenStack POC and my biggest concern was the fact that there's no real way of upgrading. Yes, you can do minor upgrades, but version changes are a nightmare. | 16:02 |
| kata-irc-bot | <lpetrut> well it can be upgraded, recent versions even support rolling upgrades - upgrading hosts one by one without having to disable the whole cloud while doing it. | 16:05 |
| kata-irc-bot | <lpetrut> can be quite easy if you're using Kolla with relatively simple deployments | 16:06 |
| kata-irc-bot | <razvan.rosca> Oh, really? Didn't tried it recently. The POC I did was about 3 years ago, it was an absolute nightmare. | 16:07 |
| kata-irc-bot | <lpetrut> things are much smoother now. I won't even mention how brittle things were 6 years ago :slightly_smiling_face: | 16:08 |
| kata-irc-bot | <razvan.rosca> Well 6 years ago I give it the first shot. It was worse than ... mmmm ... anything else. Then 5, then 4, then 3, and then I abandoned the idea. | 16:10 |
| kata-irc-bot | <razvan.rosca> Another drawback was the fact that, without the controller, the entire infra was basically dead. Did things change? Aka will networking "die" if the controller gets disconnected? | 16:10 |
| kata-irc-bot | <lpetrut> still does, but in general everything is redundant so if one service goes down, you'll have a few more keeping things up (e.g. l3 services) | 16:11 |
| kata-irc-bot | <razvan.rosca> I can remember that - when doing the POC - I removed the eth cable from the controller. It was a total chaos, internal networking was on its knees, everything was "down" (even if the kvm hosts were still running). | 16:12 |
| kata-irc-bot | <razvan.rosca> is the database redundant? I can remember that there was no way of having an automatically lb-ed DB, so I had to make a separate SQL server that had HA, and obviously failed at the first desync :P | 16:13 |
| kata-irc-bot | <lpetrut> well, you can use a db cluster, and/or separate "cells" | 16:14 |
| kata-irc-bot | <razvan.rosca> As an idea, in CloudStack and OpenNebula the Controller is needed only when managing stuff (adding/removing). Yeah, you can't obviously create new private networks with the Controller being offline. But at least everything "keeps running". | 16:14 |
| kata-irc-bot | <razvan.rosca> (also updates are at a "yum" distance) | 16:15 |
| kata-irc-bot | <lpetrut> with a proper design, you can achieve really good high availability. about upgrades, kolla is running Openstack services within containers, so upgrades will simply restart the containers with new images | 16:17 |
| kata-irc-bot | <razvan.rosca> ^ sounds interesting! | 16:17 |
| *** devimc has quit IRC | 17:17 | |
| *** devimc has joined #kata-general | 17:17 | |
| *** lpetrut has quit IRC | 17:54 | |
| *** gwhaley has quit IRC | 18:00 | |
| *** lpetrut has joined #kata-general | 18:17 | |
| *** lpetrut has quit IRC | 18:28 | |
| *** sgarzare has quit IRC | 18:44 | |
| *** devimc has quit IRC | 22:35 | |
| *** devimc has joined #kata-general | 22:46 | |
| *** devimc has quit IRC | 22:52 | |
| *** sameo has quit IRC | 23:59 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!