Post AYWaroiL0WrQc7DBZI by tuxflo@chaos.social
 (DIR) More posts by tuxflo@chaos.social
 (DIR) Post #AYUbwzBpCLyABn73CK by feld@bikeshed.party
       2023-08-07T16:20:22.816948Z
       
       1 likes, 0 repeats
       
       There's a docker-proxy process for every single port your exposeThis is crazy. Whose idea was this?
       
 (DIR) Post #AYUd6UzkPsujOLZ0y0 by mWare@mstdn.ca
       2023-08-07T16:31:26Z
       
       0 likes, 0 repeats
       
       @feld same fanboy crew that likes systemd? 16 different firewall systems? Butterfingers FS? Debates over "distributions" that really are different packaging managers for systemd? People who re-create ifconfig in a byzantine set of a dozen commands? CUPS? Route-Cache? Hiding logs? Breaking DNS? Silently using 8.8.8.8 despite that not being your preference? Need I go on? Also, I like #FreeBSD ;)
       
 (DIR) Post #AYUkFc2UIM57WZ8cwi by tuxflo@chaos.social
       2023-08-07T17:51:27Z
       
       0 likes, 0 repeats
       
       @feld for real? What will it do if you expose a port range? And how does podman deal with this?
       
 (DIR) Post #AYUkFd4IT77giSzaPA by feld@bikeshed.party
       2023-08-07T17:53:21.475903Z
       
       0 likes, 0 repeats
       
       It parses the port range and launches a process for each.at least that's what I'm seeing on my machinesno idea about podman[my opinion is that it's never worth it to use docker-proxy; use network_mode: host and move on with your life. you'll get much better performance and there are many other ways to isolate containers from each other if needed]
       
 (DIR) Post #AYV8t5dUaPXzVjnuCG by tuxflo@chaos.social
       2023-08-07T20:31:42Z
       
       0 likes, 0 repeats
       
       @feld well, I really like the idea of the overlay network. Especially when using the internal DNS to communicate between services.
       
 (DIR) Post #AYV8t8TK2gC8J44KGG by feld@bikeshed.party
       2023-08-07T22:29:25.720602Z
       
       0 likes, 0 repeats
       
       I just use localhost like always then you don't need to even use a name lolAnd really if your service gets under high load you're probably gonna end up splitting it up and moving the containers to different servers anyway
       
 (DIR) Post #AYWZORIaszDxdmb8RU by tuxflo@chaos.social
       2023-08-08T08:56:19Z
       
       0 likes, 0 repeats
       
       @feld especially when using multiple nodes the overlay network is what I prefer, because then I don't have to deal with static IP address or hostname configuration.
       
 (DIR) Post #AYWZOS0CGs6Pp0zxPE by feld@bikeshed.party
       2023-08-08T15:01:06.649148Z
       
       0 likes, 0 repeats
       
       With Swarm?
       
 (DIR) Post #AYWaDfZPaLngfetPay by tuxflo@chaos.social
       2023-08-08T15:08:20Z
       
       0 likes, 0 repeats
       
       @feld yes or stack (still don't know what means what)
       
 (DIR) Post #AYWaDgEt68yekIIXFA by feld@bikeshed.party
       2023-08-08T15:10:22.809275Z
       
       0 likes, 0 repeats
       
       both of those horrify me, so I have no plans of ever using them 🙃
       
 (DIR) Post #AYWaroiL0WrQc7DBZI by tuxflo@chaos.social
       2023-08-08T15:15:02Z
       
       0 likes, 0 repeats
       
       @feld well I have to use them and it's not that bad. But I think nowadays everyone is doing kubernetes so swam/stack are almost dead
       
 (DIR) Post #AYWarpiNHsU5iWEjGS by feld@bikeshed.party
       2023-08-08T15:17:38.517260Z
       
       0 likes, 0 repeats
       
       I also refuse to touch Kubernetes as well hahamy containers are going to continue to be managed like normal server/service deployments. The added complexity and opacity of k8s is unacceptable to me. (plus, it kills network performance with all those wasteful kernel/usespace copying of network packets... huge increase in interrupts for very little gain)
       
 (DIR) Post #AYWb1sg77jqrnuyDjc by tk@bbs.kawa-kun.com
       2023-08-08T15:19:27.865114Z
       
       0 likes, 0 repeats
       
       @feld @tuxflo Goobernetes.
       
 (DIR) Post #AYWbHPwNjm3vJEzJI0 by tuxflo@chaos.social
       2023-08-08T15:20:16Z
       
       0 likes, 0 repeats
       
       @feld always depends on your usecase/needs. If you want some "high availability" or "rolling, downtimeless updates" then I think it's better to deal with k8s right away and don't try to build your own custom solution.
       
 (DIR) Post #AYWbHQfP2O4hYs3GSm by feld@bikeshed.party
       2023-08-08T15:22:16.035440Z
       
       0 likes, 0 repeats
       
       the real problem is that too many project managers think everything needs "rolling, downtimeless updates"
       
 (DIR) Post #AYWbMQieGorrAhxEYq by tuxflo@chaos.social
       2023-08-08T15:23:16Z
       
       0 likes, 0 repeats
       
       @feld yes, full ack!
       
 (DIR) Post #AYWbn9P6kunVHwYWeW by tuxflo@chaos.social
       2023-08-08T15:25:01Z
       
       0 likes, 0 repeats
       
       @feld But only then you can tick the "Cloud Ready" Buzzword Bingo card
       
 (DIR) Post #AYWbnASKqOyOYF4cK0 by feld@bikeshed.party
       2023-08-08T15:27:45.732295Z
       
       0 likes, 0 repeats
       
       it's all peacocking so the PM can look good to the CIO or something even though it does nothing but check that box and increase complexity and costs of running the service (resources -- computing and human engineering)