Hi

I understand the convenience of NFS, however in NFS, filesystem is at
Storage end, and the CPU , RAM  performance of Storage become very very
important .  The Overload  may happen on storage rather
compute/hybervisor.

Where as iscsi based sharedmountpoint might have more efficiency for this.

I am facing issue is ,  i create first host with sharemoutpint /vmdisk ,
 there is no issue of this,  After that i add another host the the Cluster,
and and I mount the same name of ShareMountPoint (/vmdiak) in the newly
added host ,  since /vmdisk already appear in Primary Storage , and is in
"Cluster" , so I did not create it again .
I assume this shall work but apparently the /vmdisk in newly added in
hypervisor host is not recognizable. .




Thank you


On Fri, Sep 11, 2020 at 8:45 PM Andrija Panic <andrija.pa...@gmail.com>
wrote:

> Hi Hean,
>
> this is a SharedMountPoint option when adding Primary Storage in
> CloudStack.
>
> This does require that your 2 hosts (not sure you'll be able to do it only
> with 2 hosts) have some kind of Clustered Filesystem like clvm, GFS2, OCFS
> set up on that mount point /LUN  - none of those I would recommend really
> due to long-term stability etc - if you, can avoid it, if you can't avoid
> it, then... use it.
>
> The /vmdisk folder will be the sharedMountPoint and since more than one
> compute hosts are writing to a shared LUN this is why you need a clustered
> filesystem on top of that LUN.
>
> I would, instead, advise you to explore simple NFS setup or similar.
>
> best,
>
> On Fri, 11 Sep 2020 at 07:12, Hean Seng <heans...@gmail.com> wrote:
>
> > Hi
> >
> > Is anybody know One Cluster have two KVM host, and each host has same
> > ShareMountPoint Name,
> >
> > And I to create one Primary Storage with same SharedMountPoint Name,  for
> > example, /vmdisk on both host . (both in same cluster)
> >
> > And when create VM,  it will automatically create the vm-storage-disk
> > inside the Host's Sharemount point .
> >
> >
> > --
> > Regards,
> > Hean Seng
> >
>
>
> --
>
> Andrija Panić
>


-- 
Regards,
Hean Seng

Reply via email to