[lxc-users] LXD - Production Hardware Guide

Steven Spencer sspencerwire at gmail.com
Sun Jun 7 16:05:34 UTC 2020


Thanks Stéphane!

On Fri, Jun 5, 2020 at 9:42 PM Stéphane Graber <stgraber at stgraber.org>
wrote:

> ZFS very much prefers seeing the individual physical disks.
> When dealing directly with full disks, it can do quite a bit to
> monitor them and handle issues.
> If it's placed behind RAID, whether hardware or software backed, all
> that information disappears and it doesn't really know whether to
> retry an operation or consider the disk to be bad.
>
> The same is true if you ever end up using Ceph where you want a 1:1
> mapping between physical disk and OSDs, so in general I'd recommend
> against hardware RAID at this point.
>
> Stéphane
>
> On Fri, Jun 5, 2020 at 4:59 PM Steven Spencer <sspencerwire at gmail.com>
> wrote:
> >
> > Andrey and List,
> >
> > Thanks so much for your response, and we understand all of that. We know
> that if we have 3 containers the server configuration is going to be
> different than if we have 30 or 100, and that we will have to size RAM,
> Processors, etc, accordingly. What I think we are more interested in is: If
> we use ZFS, is there a recommended way to use it? Should we use RAID of any
> kind? If so, should it be hardware or software RAID? We realize, too, that
> we will need to size our drives according to how much space we will
> actually need for the number of containers we will be running. Really it's
> just about the underlying file system for the containers. It seems like
> there should be a basic white paper or something with just guidelines on
> best practices for LXD. That would really help us. We have found the LXD
> documentation and we have actually used these docs. We've even used ZFS
> under LXD on our first iteration of this project about 3 years ago. We are
> now looking to do this again. The first time was mostly a success. Recenly,
> we had the main LXD server die and for no apparent reason (hardware /
> software / memory - the logs don't really give us much). Our snapshot
> server was the savior, but now we need to repeat our earlier process, and
> if we made mistakes, we would like to fix them in the process.
> >
> > Thanks again for the response, any further information would be helpful.
> >
> > Steven G. Spencer
> >
> >
> >
> > On Fri, Jun 5, 2020 at 2:20 PM Andrey Repin <anrdaemon at yandex.ru> wrote:
> >>
> >> Greetings, Steven Spencer!
> >>
> >> > Is there a good link to use for specifying hardware requirements for
> an LXD dedicated server?
> >>
> >> There can't be specific requirements, it all depends on what you want
> to do,
> >> how many containers to run, etc.
> >>
> >>
> >> --
> >> With best regards,
> >> Andrey Repin
> >> Friday, June 5, 2020 22:05:54
> >>
> >> Sorry for my terrible english...
> >>
> >> _______________________________________________
> >> lxc-users mailing list
> >> lxc-users at lists.linuxcontainers.org
> >> http://lists.linuxcontainers.org/listinfo/lxc-users
> >
> > _______________________________________________
> > lxc-users mailing list
> > lxc-users at lists.linuxcontainers.org
> > http://lists.linuxcontainers.org/listinfo/lxc-users
>
>
>
> --
> Stéphane
> _______________________________________________
> lxc-users mailing list
> lxc-users at lists.linuxcontainers.org
> http://lists.linuxcontainers.org/listinfo/lxc-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxcontainers.org/pipermail/lxc-users/attachments/20200607/10f0bfaa/attachment.htm>


More information about the lxc-users mailing list