Great talk. I want to emphasize that any attempt to introduce a platform comes with an added cost of maintaining that platform. When the platform, and the solutions on top of it, have to be hosted on-premises, it is crucial to keep this cost in mind since a team may not have the luxury of delegating maintenance to someone else.
Sad story about what happened to IT industry. We can't afford anything, we have to do it immediately in 'fire mode', it has to work somehow, go, tik tok, tik tok, why is it taking so long, tik tok... Everything is one giant MVP nowadays. Unfortunately its trendy to solve obvious problems on production, instead of anticipating them. Culture of non-thinking. Culture of empirically head-bumping everything.
Great talk. I am thinking about replace Kafka with NATS. Furthermore I see the HUMIO it is a good replacement ( memory consumption and bootup time for example ) for elasticsearch ?
As he said at the start, HA isnt the target here, i would trade docker-compose/swarm with a lightweight kubernetes any day. Just think about configuring all the stuff that base-kubernetes comes with in a docker-compose deployment
A special case of clusters where number of nodes is 1. Still make sense because. It's very easy to scale up by adding nodes when need arises. It not ha while it is single node. That's all
Obviously not an appropriate production deployment for most systems, but that's not the interesting part. The interesting part to me is seeing the progress in "scaling down" k8s so it's not this enterprise-only juggernaut anymore. Running k8s in this (obviously not ideal) setup should smooth the transition to a proper scalable, fault-tolerant setup down the road; compared with the more straightforward approach of building automation scripts around his manual steps, which would create more and more team-owned complexity and technical debt, and move their solution AWAY from best practices rather than toward them.
Great talk. A real engineer here with that brutal problem solving attitude, superb. 😊
Great talk. I want to emphasize that any attempt to introduce a platform comes with an added cost of maintaining that platform. When the platform, and the solutions on top of it, have to be hosted on-premises, it is crucial to keep this cost in mind since a team may not have the luxury of delegating maintenance to someone else.
Sad story about what happened to IT industry. We can't afford anything, we have to do it immediately in 'fire mode', it has to work somehow, go, tik tok, tik tok, why is it taking so long, tik tok... Everything is one giant MVP nowadays. Unfortunately its trendy to solve obvious problems on production, instead of anticipating them. Culture of non-thinking. Culture of empirically head-bumping everything.
This.
@@derskeal4240 yddda
Great insights! Thank you!)
Informative.Thanks for this talk.
Great talk. I am thinking about replace Kafka with NATS. Furthermore I see the HUMIO it is a good replacement ( memory consumption and bootup time for example ) for elasticsearch ?
Humio is a great choice for logs. You may also want to take a look at Loki from Grafana Labs.
@@JonasKongslund I have it on my list ;-)
I wouldn’t consider NATS and Kafka the same, one is MQ and more is event source / write ahead log.
Sorry, what does this have to do with the video?
Why not just use Docker swarm?
The repo is not public
why not use podman ?
This doesn't really make sense; replace docker compose with Kubernetes - complicating everything. It isn't even highly available!
As he said at the start, HA isnt the target here,
i would trade docker-compose/swarm with a lightweight kubernetes any day. Just think about configuring all the stuff that base-kubernetes comes with in a docker-compose deployment
Great video, thx :)
Single-node and cluster?
Haha
A special case of clusters where number of nodes is 1. Still make sense because. It's very easy to scale up by adding nodes when need arises. It not ha while it is single node. That's all
1 all
Hard pass but thanks
Obviously not an appropriate production deployment for most systems, but that's not the interesting part. The interesting part to me is seeing the progress in "scaling down" k8s so it's not this enterprise-only juggernaut anymore. Running k8s in this (obviously not ideal) setup should smooth the transition to a proper scalable, fault-tolerant setup down the road; compared with the more straightforward approach of building automation scripts around his manual steps, which would create more and more team-owned complexity and technical debt, and move their solution AWAY from best practices rather than toward them.
🤬