Re: [Moosefs-users] Virtual machine storage

[Elliot F. <efi...@gm...>]

Give it enough spindles to handle the IO load that you're throwing at it
and it works just fine.  VMs going into RO mode is because of lengthy IO
wait - i.e. not enough resources for what you want to do.

On Mon, Oct 27, 2014 at 11:49 AM, WK <wk...@bn...> wrote:

> We tried it.
>
> Worked great until you had situations where the cluster was ALSO
> replicating and/or was under load.
>
> Then the VMs would start going Read-Only.
>
> We love MooseFS, but for VMs we use DRBD or GlusterFS with simple
> replication.
>
>
> On 10/23/2014 2:01 PM, Joseph Love wrote:
> > Hi,
> >
> > I’m curious if anyone has used MFS as a virtual machine storage system
> (specifically, as the storage for a VMware environment).  Last tidbits I
> read suggested that the latencies were a bit too high to use MFS in this
> way, and instead to use it as the storage *in* the virtual machines instead.
> >
> > Some testing with a 3-node cluster and using bonnie++, iozone, &
> tiobench suggests that write performance for small, random blocks suffer
> from poor performance and high latencies.  If i’m not mistaken, that should
> be pretty similar to what VMware disk updates act like, so it seems like
> it’s possibly a bad idea.  Bigger blocks, and large sequential data access
> seem to perform great, though.
> >
> > So, anyone with real-world experience with using MFS in this way?
> >
> > Thanks,
> > -Joe
> >
> >
> >
> ------------------------------------------------------------------------------
> > _______________________________________________
> > moosefs-users mailing list
> > moo...@li...
> > https://lists.sourceforge.net/lists/listinfo/moosefs-users
>
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> moosefs-users mailing list
> moo...@li...
> https://lists.sourceforge.net/lists/listinfo/moosefs-users
>