To add some extra information regarding this matter; it is not just a
matter of ACS supporting other technologies such as Gluster; hypervisors
need to support the storage type in order to enable ACS to expose its
configuration to users.

On Fri, Nov 17, 2017 at 12:26 PM, Rubens Malheiro <[email protected]
> wrote:

> I believe the best solution for replication is GlusterFS, I have a working
> although I can never make cloudstack operate natively I use glusterfs and
> nfs-ganesha
> I'm sorry, my english
>
> On Fri, Nov 17, 2017 at 7:48 AM, William Alianto <[email protected]>
> wrote:
>
> > Hi,
> >
> > Sorry for the late reply. So I summarized that NFS with DRDB is enough
> for
> > POC purpose but not recommended for large scale production, where ceph
> > would be more suitable.
> >
> > Thanks for the insight for this problem.
> >
> > --
> > Regards,
> >
> > William
> >
> >
> > On 09-Nov-17 19:28:49, Ivan Kudryavtsev <[email protected]>
> wrote:
> > Hi. I wouldn't recommend go with HA NFS, because you have to use it in
> sync
> > mode which leads to poor performance.
> >
> > I suggest using ceph with ssd (or even optane for log devices), because
> > it's the thing which is supposed to be clustered and kvm "knows" how to
> use
> > it properly and handle node outages via librados.
> >
> > But, still, nfs sync over drbd and keepalived should work, but not a
> > scalable solution, of course.
> >
> > 9 нояб. 2017 г. 19:10 пользователь "Makrand"
> > написал:
> >
> > > Hi William,
> > >
> > > The HA capabilities for storage are offered by OS/Vendor you're using
> for
> > > storage. Different projects use a different solution for storage HA.
> (HA
> > > mostly a paid/premium affair). e.g. for one of the zone, we are using
> > LUNs
> > > coming from Hitachi storage array. Hitachi array is the
> enterprise-class
> > > thing which has HA built in.
> > >
> > > If you're just starting and want to test things. I would recommend
> using
> > > something like FreeNAS. This software-defined storage that caters ZFS.
> > ZFS
> > > offers a software RAID and has lots of features. Plus you can get both
> > > NFS/iSCSI.
> > >
> > > If you're having time in hand, try DRBD (DRBD just replicates the
> blocks
> > > from disk to disk over the network). You create highly available NFS
> (ACS
> > > supports both primary and sec on NFS. Just make sure you have
> sufficient
> > > network bandwidth between hypervisor hosts and storage system) cluster
> > > using DRBD on Linux. This would be a bit time consuming to set up.
> > >
> > >
> > >
> > >
> > > --
> > > Makrand
> > >
> > >
> > > On Thu, Nov 9, 2017 at 9:41 AM, William Alianto wrote:
> > >
> > > > Hi,
> > > >
> > > > I'm currently still learning about ACS deployment. I have seen about
> > the
> > > > solution for HA management server and hypervisor, but I still don't
> > know
> > > > how to create HA storage cluster for Primary and Secondary storage. I
> > > would
> > > > think of making HA NFS storage using HA proxy or maybe using
> something
> > > like
> > > > Ceph Cluster. Would there be any better option for the HA storage
> > > solution?
> > > >
> > > > --
> > > > Regards,
> > > >
> > > > William
> > >
> >
>



-- 
Rafael Weingärtner

Reply via email to