4 ms·

Can anyone post an experience using rkt [0]. I'd love an alternative to using Docker for containers, though I'm not sure how well a community has evolved around

by ghayes 8y ago

Can anyone post an experience using rkt [0]. I'd love an alternative to using Docker for containers, though I'm not sure how well a community has evolved around rkt. [0] https://github.com/rkt/rkt https://github.com/rkt/rkt

edoceo 8y ago
This comment made/makes me steer away https://github.com/rkt/rkt/issues/3912 https://github.com/rkt/rkt/issues/3912 There is another open issue claiming the project is dead (but maybe it's just stable) My whole team is onboard with docker, and we don't need account to install/use.
- ori_b 8y ago
  Why do you believe that churn is a good thing, especially for a thin wrapper around Linux namespaces? There should not be much this is actually doing, so lots of commits would be worrying.
  - edoceo 8y ago
    I didn't say that that churn is a good thing. The github-issue and a sibling are folks asking this question. Why? Lingering bugs? Waiting for features? That the question is open is the concern. And again, compared to my existing tooling which is not having that question and has momentum. Evaluated against docker (a recent move for us) but didn't choose. We also looked a "roll your own" type solutions - required expertise our team didn't have.
- shakna 8y ago
  Last commit 19 days ago. Code commits merged into master seem to be around every 3 months for the last couple years. That seems healthy to me. As for documentation, it seems to have monthly or bimonthly updates. The question in your issue is that their is a lack of activity... Came about three weeks after a commit, and two weeks before another. The issue just seems to be that rkt prefers their mailing list to GitHub issues, though they do pay attention to both. And I don't see a problem with that.
solarengineer 8y ago
Please check the Open Containers Initiative [0]. That lists a number of runtimes. There are now entire tool chains developed independent of anything from Docker Inc. [0] https://www.opencontainers.org https://www.opencontainers.org
alpb 8y ago
Docker for Desktop (which is what the issue text describes) is not the same thing as rkt so it would not be an apples to apples comparison. You can still install docker with "apt-get" on Linux. https://docs.docker.com/install/linux/docker-ce/ubuntu/ https://docs.docker.com/install/linux/docker-ce/ubuntu/ Docker for Desktop is the GUI experience that runs the docker engine in a VM on your machine with some port forwarding magic etc. rkt equivalent would be containerd probably.
- paulie_a 8y ago
  Honestly docker apt-get previously sucked to the extent I wouldn't bother attempting again.
  - deadbunny 8y ago
    Adding a repo and running `apt-get update && apt-get install docker-ce` sucks?
    - paulie_a 8y ago
      Yes when it needs tweaking on every update and repeatedly fails. It's obnoxious. It was immature technology in my personal experience. Maybe others have had a better experience but I was turned off by numerous issues.
ilovecaching 8y ago
If your goal is to just use containers, you don't need Docker. You can use containerd, lxc, nspawn, or use clone directly. Docker is an extremely bloated piece of software, and you can probably get by with just a little knowledge of Linux APIs and a solid understanding of your project's requirements.
colemickens 8y ago
I'm surprised to see rkt mentioned so much. This thread has a very 2017 vibe and doesn't seem to account for m(any) of the changes in Kubernetes (CRI), docker/containerd, anything with regard to what RH is doing as they bet more and more on this space. Anyway... Red Hat's crio is a much more obvious contender in this space (my issues with RH "marketing" of crio vs containerd aside). A slightly different scope, but even Kata's future seems more promising than rkt's. Or, just use containerd if you just want a clearly OSS engine with the weight of docker behind it. ('containerd' is Docker's container engine, effectively)
- conradfr 8y ago
  I cant believe people are still taking and using technologies from 2017 /s To be fair this thread is also a testament to the sad state of affair in the container industry. How can you even chose a technology confidently? It seems even more madness than the JS world. Sometimes I feel like a dinosaur with my Vagrant boxes and my dedicated server. With threads like this I feel great.
  - colemickens 8y ago
    Rkt didn't get much attention in 2017 at all. I only said that because 'rkt' was often portrayed as the anti-docker years ago, and it's really no longer a good way to look at the playing field. It wasn't about making fun of rkt or anything. Also, I don't know, CRI has been in the works for ages at this point, as well as crio. It just seems like people who aren't paying attention in this space remember HN threads from 18-24 months ago where people thought rkt was going to be the savior from docker. It's just pretty out of date. The comment about vagrant and a dedicated server embody an attitude that is just depressing to me. Yes, if you stick with old, functional technologies, you won't need to pay attention and/or learn anything new. And that's fine, if you're happy with it. There's a reason the rest of us moved on. Learning about CRI and knowing that rkt is dead and crio is the alternative to containerd is not that crazy of knowledge to have, or maybe I'm underestimating how much I know in the k8s space.
raesene9 8y ago
The slight concern I would have with rkt, alongside some other of CoreOS' projects is that their future Post RedHat acquisition isn't super clear. In the case of rkt, AFAIK Redhat's container runtime (Cri-o) uses containerd and runc under the covers, and I've not seen any indication that they'll change that, so not too sure where rkt fits in that landscape.

by ghayes 8y ago

edoceo 8y ago

This comment made/makes me steer away https://github.com/rkt/rkt/issues/3912 https://github.com/rkt/rkt/issues/3912 There is another open issue claiming the project is dead (but maybe it's just stable) My whole team is onboard with docker, and we don't need account to install/use.

ori_b 8y ago
Why do you believe that churn is a good thing, especially for a thin wrapper around Linux namespaces? There should not be much this is actually doing, so lots of commits would be worrying.
- edoceo 8y ago
  I didn't say that that churn is a good thing. The github-issue and a sibling are folks asking this question. Why? Lingering bugs? Waiting for features? That the question is open is the concern. And again, compared to my existing tooling which is not having that question and has momentum. Evaluated against docker (a recent move for us) but didn't choose. We also looked a "roll your own" type solutions - required expertise our team didn't have.
shakna 8y ago
Last commit 19 days ago. Code commits merged into master seem to be around every 3 months for the last couple years. That seems healthy to me. As for documentation, it seems to have monthly or bimonthly updates. The question in your issue is that their is a lack of activity... Came about three weeks after a commit, and two weeks before another. The issue just seems to be that rkt prefers their mailing list to GitHub issues, though they do pay attention to both. And I don't see a problem with that.

solarengineer 8y ago

Please check the Open Containers Initiative [0]. That lists a number of runtimes. There are now entire tool chains developed independent of anything from Docker Inc. [0] https://www.opencontainers.org https://www.opencontainers.org

alpb 8y ago

Docker for Desktop (which is what the issue text describes) is not the same thing as rkt so it would not be an apples to apples comparison. You can still install docker with "apt-get" on Linux. https://docs.docker.com/install/linux/docker-ce/ubuntu/ https://docs.docker.com/install/linux/docker-ce/ubuntu/ Docker for Desktop is the GUI experience that runs the docker engine in a VM on your machine with some port forwarding magic etc. rkt equivalent would be containerd probably.

paulie_a 8y ago
Honestly docker apt-get previously sucked to the extent I wouldn't bother attempting again.
- deadbunny 8y ago
  Adding a repo and running `apt-get update && apt-get install docker-ce` sucks?
  - paulie_a 8y ago
    Yes when it needs tweaking on every update and repeatedly fails. It's obnoxious. It was immature technology in my personal experience. Maybe others have had a better experience but I was turned off by numerous issues.

ilovecaching 8y ago

If your goal is to just use containers, you don't need Docker. You can use containerd, lxc, nspawn, or use clone directly. Docker is an extremely bloated piece of software, and you can probably get by with just a little knowledge of Linux APIs and a solid understanding of your project's requirements.

colemickens 8y ago

I'm surprised to see rkt mentioned so much. This thread has a very 2017 vibe and doesn't seem to account for m(any) of the changes in Kubernetes (CRI), docker/containerd, anything with regard to what RH is doing as they bet more and more on this space. Anyway... Red Hat's crio is a much more obvious contender in this space (my issues with RH "marketing" of crio vs containerd aside). A slightly different scope, but even Kata's future seems more promising than rkt's. Or, just use containerd if you just want a clearly OSS engine with the weight of docker behind it. ('containerd' is Docker's container engine, effectively)

conradfr 8y ago
I cant believe people are still taking and using technologies from 2017 /s To be fair this thread is also a testament to the sad state of affair in the container industry. How can you even chose a technology confidently? It seems even more madness than the JS world. Sometimes I feel like a dinosaur with my Vagrant boxes and my dedicated server. With threads like this I feel great.
- colemickens 8y ago
  Rkt didn't get much attention in 2017 at all. I only said that because 'rkt' was often portrayed as the anti-docker years ago, and it's really no longer a good way to look at the playing field. It wasn't about making fun of rkt or anything. Also, I don't know, CRI has been in the works for ages at this point, as well as crio. It just seems like people who aren't paying attention in this space remember HN threads from 18-24 months ago where people thought rkt was going to be the savior from docker. It's just pretty out of date. The comment about vagrant and a dedicated server embody an attitude that is just depressing to me. Yes, if you stick with old, functional technologies, you won't need to pay attention and/or learn anything new. And that's fine, if you're happy with it. There's a reason the rest of us moved on. Learning about CRI and knowing that rkt is dead and crio is the alternative to containerd is not that crazy of knowledge to have, or maybe I'm underestimating how much I know in the k8s space.

raesene9 8y ago

The slight concern I would have with rkt, alongside some other of CoreOS' projects is that their future Post RedHat acquisition isn't super clear. In the case of rkt, AFAIK Redhat's container runtime (Cri-o) uses containerd and runc under the covers, and I've not seen any indication that they'll change that, so not too sure where rkt fits in that landscape.