Sorry for replying late, I was on vacation since 24Jan.
On 2014年01月24日 11:29, Aline Manera wrote:
On 01/22/2014 11:43 AM, Royce Lv wrote:


Guys,

    When testing with kimchi nfs feature, I filed two issues related to
nfs image permission problem:

    1. volume creation failure: Because of storage pool permission is
not configured to make write
permission.(https://github.com/kimchi-project/kimchi/issues/261)
    2. vm with volume cannot be started: Root users are mapped to
nobody, so img it created cannot be accessed by libvirt-qemu(on ubuntu)
user.(https://github.com/kimchi-project/kimchi/issues/267)

    After discussed with Mark Wu, we would like to propose the
following to resolve these two problem:

    1. To allow creation: export with all_squash(gid = kimchi_guid) and
group allow write permission. Also with planned nfs-pool prevalidation
(a timeout try mount in a process), we can check if the gid and
permission is right. This will save us from future trouble.


It is only related to NFS server setup, right?

/home/alinefm *(rw,all_squash,anongid=<kimchi-gid>)

So for that we should only write instructions to user on README or
other doc file.
True

    2. To allow qemu process(started by libvirt) to access img, we add
uid ('qemu' under fedora and 'libvirt-qemu' under ubuntu) which running
qemu process to 'kimchi' group to allow the write access of the img.


For that we need to create a group with the same kimchi gid used for NFS server
and then add 'qemu/libvirt-qemu' to this group?

Is that right?
Right, and we validate when mount a nfs pool to check if the permission is right.

    I am also investigating other possibilities like using storage pool
permissions and so on.
    Welcome thoughts on it!