|
pram posted:like for example doing something as simple as saving images to a disk, in a clustered setup, is not possible. you now need an object store or some distributed storage. it ratchets up complexity a ton. most people don't need to deploy infra like this this is actually really easy and idg why people bring it up so much probably most of your app doesnt need storage, okay, so thats easy, throw in stateless containers and then like some things do (like your db), so make them a daemonset in k8s and use node storage, okay that was easy problems left are: 'there's no distributed object store' which is also easy to fix if you install a distributed object store or use nfs which probably sucks, or ceph or gluster or whatever that other one that had the company implode and pvcs, or use k8s in the cloud like you probably already are and use your cloud providers block storage
|
# ¿ Jun 8, 2017 19:24 |
|
|
# ¿ May 14, 2024 19:04 |
|
containers are rad for production, they're just the best iteration of 30(?) years of jails with the best generic tooling so far except docker is utter trash, like the specific software is poopoo
|
# ¿ Jun 8, 2017 19:27 |
|
i'm sorry i mean moby, not docker, pbuh
|
# ¿ Jun 8, 2017 19:28 |
|
Doom Mathematic posted:Does "Write once, run anywhere" still come with the caveat "as long as it's 64-bit Linux"? well more importantly, write once run anywhere doesn't actually work anywhere else that promises it and the caveat is 'with whatever architecture you built the container with' but lmbo if you're running production that isn't 64-bit linux
|
# ¿ Jun 11, 2017 16:25 |
|
Visual GNUdio posted:Here is what is cool and good about containers: it forces developers to actually state their loving requirements and put all that poo poo together so I don't have to janitor their special snowflakes, while keeping their stupid fingers far away from the infrastructure. mega
|
# ¿ Jun 11, 2017 22:13 |