Simple questions about server use case, ZFS, Qemu, iptables

General Discussion
#1

Hi,

I am very new with the CL but the first impression was promising of the philosophy and design. I have read tutorial about a ZFS, a CL documentation and some community threads about CL in a general. Some questions still pop-up.

I am trying to so choose a general server distribution for the ZFS disk-management ( OpenZFS 2.0 ) and the qemu hypervisor for relay on zfs-pools. That is about all I need. The root on a normal mdadm and the iptables of course.

  1. Is that the CL is stable enough? I have read about some boot issues after updates and this is not acceptable in a server use. Is manual updating really the best option to keep system updating with a delay and waiting for that bugs will be solved from the “first wave of updates”? Is there option to come a one update behind? Or just the custom service to update rarely enough?

  2. How often a LTS kernel will be updated? Is that following the mainline LTS system? If I understood a mixer right. It is possible to compile and package to the bundle the ZFS in the way that relays on the specific kernel version and if a newer kernel comes that bundle prevent whole system to update? For example ZFS 2.0 officially support 5.9 and the next LTS will be soon 5.10 so that is the problem from the first day if there is no a known schedule.

  3. Will it be a problem if a system is unbooted for long period of time. That there is too long cap between the running and the updated kernel and other parts of the system? Will services still be updated without booting?

Tell me right away if I am trying to seek my goals from a completely wrong place :slight_smile:

#2
  1. CL is on rolling release. In general a distro can not say it both has latest packages and great stability. So I don’t think CL is “stable” in the sense of Debian’s stability.
  2. Kernels are updated once it’s released, usually. So you can refer to the update schedule of LTS kernels.
  3. CL uses delta update. Similar to how git work, if you are behind the HEAD of a branch for many commits, the update will need to download lots of stuff and is slow.

Well, if you need stability for server, and want good support, I recommend BSD or RedHat. BSD is quite stable, or rather I should say notoriously stable that some servers are running for decades without restart. RedHat has great customer support.

1 Like
#3

Then it comes to a question why CL, which aims at cloud/server, has rolling release.

Nowadays softwares have short release cycle, and sometimes there are major feature update every several months. If the official repo doesn’t update frequently, then system administrators end up compiling stuff by themselves, that’s not fun and lost the stability guaranteed by the repo maintainer.

Also, rolling release is a way to automatically mitigate possible attack surface by constantly receiving bug fixes/patches. The could be crucial in certain use cases.

Sure, rolling release by nature is not stable. Any one update may has stuff that break a certain component your server needs. But:

  • swupd make it possible to rollback to previous OS version
  • Issues at a specific OS version means nothing when it’s compared to long-run overall benefits
1 Like
#4

Clear linux as a server is quite stable. Especially if you run your own updates on the weekends or something. I recommend ansible for this.

I wouldn’t recommend CL for ZFS application unless you really have been doing this for years and have been practicing rebuilding kernels and modules and installing zfs since zols early days. Better to use Debian or Ubuntu.

Also as @doct0rHu suggested, Freebsd is very stable (so is Debian) and Freebsd supports zfs out of the box. Of course Debian is much better for a hypervisor with kvm/qemu/libvirt.

3 Likes