is it practical to replace pve's lxc totally with docker?

Comments

• minlearn Member

  September 2024

  omg
  why i posted in provider cat?
  please move it to general cat
  thank you
• lukast__ Member, Megathread Squad

  September 2024

  While LXC and Docker both use Linux namespaces and cgroups, they're quite different as LXC contains a complete distribution with a startup system and often sshd, but Docker (AFAIK) only has the daemons.
  And the root filesystems are also different, IIRC Docker's root filesystems are readonly.
  (If I'm wrong feel free to correct me, never used Docker and never will.)
• minlearn Member

  September 2024

  @lukast__ said:
  While LXC and Docker both use Linux namespaces and cgroups, they're quite different as LXC contains a complete distribution with a startup system and often sshd, but Docker (AFAIK) only has the daemons.
  And the root filesystems are also different, IIRC Docker's root filesystems are readonly.
  (If I'm wrong feel free to correct me, never used Docker and never will.)

  precisely
• minlearn Member

  September 2024 edited September 2024

  @lukast__ said:
  While LXC and Docker both use Linux namespaces and cgroups, they're quite different as LXC contains a complete distribution with a startup system and often sshd, but Docker (AFAIK) only has the daemons.
  And the root filesystems are also different, IIRC Docker's root filesystems are readonly.
  (If I'm wrong feel free to correct me, never used Docker and never will.)

  you are quite right

  just because lxc relies on emulating linux rootfs details , it turned out too heavy
  docker however，just lit itself upper a bit beyond kernel, so it turned out light and to be fit for "sandbox runtime"

  in fact, docker implement has wider applicable areas and domains than lxc
• lukast__ Member, Megathread Squad

  September 2024

  @minlearn said: just because lxc relies on emulating linux rootfs details , it turned out too heavy

  No? What about LXC is "too heavy"?

  @minlearn said: docker however，just lit itself upper a bit beyond kernel, so it turned out light and to be fit for "sandbox runtime"

  Docker is by no means "light". Just checked the source code:

  $ find . -name \*.go -exec wc -l {} \; | sed 's/ .*//' | awk '{s+=$1} END {print s}'
  1949259
  

  Nearly 2M LoC!
• emgh Member, Megathread Squad

  September 2024

  I love Docker
• darkimmortal Member

  September 2024

  systemd-nspawn is where it’s at if you want a halfway sane container system
• minlearn Member

  September 2024 edited September 2024

  @lukast__ said:

  @minlearn said: just because lxc relies on emulating linux rootfs details , it turned out too heavy

  No? What about LXC is "too heavy"?

  @minlearn said: docker however，just lit itself upper a bit beyond kernel, so it turned out light and to be fit for "sandbox runtime"

  Docker is by no means "light". Just checked the source code:

  $ find . -name \*.go -exec wc -l {} \; | sed 's/ .*//' | awk '{s+=$1} END {print s}'
  1949259
  

  Nearly 2M LoC!

  while sharing the same kernel and namespaces/cgroups basis:

  docker requires no systemd, no init in rootfs, just a executable path as daemon

  but lxc rootfs requires explict init mechanism and systemd/systemd alternatives just like normal linux rootfs，services rooted in systemd can be a very huge network

  docker has 2M LoC codebase, mostly of them are transaction level api, not implement level logics, the basic parts are still thin as lxc.
• lukast__ Member, Megathread Squad

  September 2024

  @minlearn said: but lxc rootfs requires explict init mechanism and systemd/systemd alternatives just like normal linux rootfs，services rooted in systemd can be a very huge network

  From https://linuxcontainers.org/lxc/manpages/man5/lxc.container.conf.5.html:

  lxc.init.cmd
     Absolute path from container rootfs to the binary to use as init. This mostly makes sense for lxc-start. Default is /sbin/init.