[Gluster-users] nfs-ganesha/pnfs read/write path on EC volume

[Jiffin Tony Thottan]

On 18/03/16 11:04, Serkan Çoban wrote:
> I am using 2.3. But I can test 2.4 if there are centos packages..
> Or I can try to build rpm packages from latest ganesha if it is easy to build..
Unfortunately I can't find any epel/centos[1] packages for 2.4, its only 
for fedora24[2] packages
CCing Kaleb who knows about next 2.3.x update.

Following are the steps  used in ci.centos.org to build rpms(I hope it 
works)

# install NFS-Ganesha build dependencies
yum -y install git bison flex cmake gcc-c++ libacl-devel krb5-devel 
dbus-devel libnfsidmap-devel libwbclient-devel libcap-devel 
libblkid-devel rpm-build redhat-rpm-config

# install the latest version of gluster
yum -y install centos-release-gluster
yum -y install glusterfs-api-devel

git init nfs-ganesha
cd nfs-ganesha
git fetch https://review.gerrithub.io/ffilz/nfs-ganesha next

git checkout -b next FETCH_HEAD

# update libntirpc
git submodule update --init || git submodule sync

mkdir build
cd build

cmake -DCMAKE_BUILD_TYPE=Maintainer ../src && make rpm

[1] http://mirror.centos.org/centos-7/7.2.1511/storage/x86_64/gluster-3.7/
[2] http://arm.koji.fedoraproject.org/koji/packageinfo?packageID=17239

--
Jiffin
> On Fri, Mar 18, 2016 at 7:03 AM, Jiffin Tony Thottan
> <jthottan at redhat.com> wrote:
>>
>> On 17/03/16 23:17, Serkan Çoban wrote:
>>> Hi Jiffin,
>>> Will these patches land in 3.7.9?
>>
>> Hi Serkan,
>>
>> I moved all changes to ganesha [1] and got merged upstream (ganesha
>> V2.4-dev-9),
>> I missed to back port it to 2.3, so which ganesha build are u using?
>>
>> [1] https://review.gerrithub.io/#/c/263180/
>>
>> Jiffin
>>
>>> Serkan
>>>
>>> On Tue, Feb 16, 2016 at 2:47 PM, Jiffin Tony Thottan
>>> <jthottan at redhat.com> wrote:
>>>> Hi Serkan,
>>>>
>>>> I had moved out previous gfapi-side to ganesha and include all those
>>>> change
>>>> in single patch https://review.gerrithub.io/#/c/263180/
>>>>
>>>> I will try get it reviewed and merge the patch as soon as possible.
>>>>
>>>> With Regards,
>>>> Jiffin
>>>>
>>>> On 14/02/16 21:54, Serkan Çoban wrote:
>>>>> Thanks for the answer,
>>>>> AFAIK, when using pNFS, every different file read/write should go to
>>>>> different server in order to utilize all servers in parallel.
>>>>> I am waiting for the patches and future releases.Thanks for all your
>>>>> efforts.
>>>>>
>>>>> Serkan
>>>>>
>>>>> On Wed, Feb 10, 2016 at 2:21 PM, Jiffin Tony Thottan
>>>>> <jthottan at redhat.com> wrote:
>>>>>> Hi,
>>>>>>
>>>>>> Sorry for the delayed delayed response
>>>>>>
>>>>>>
>>>>>> On 10/02/16 13:51, Pranith Kumar Karampuri wrote:
>>>>>>>
>>>>>>>
>>>>>>> On 02/10/2016 01:15 PM, Serkan Çoban wrote:
>>>>>>>> Hi Jiffin,
>>>>>>>>
>>>>>>>> Any update about the write path?
>>>>>>> I saw him send some mails related to this, yesterday and day before.
>>>>>>> You
>>>>>>> will hear from him soon.
>>>>>>>
>>>>>>> Pranith
>>>>>>>>
>>>>>>>> Serkan
>>>>>>>>
>>>>>>>> On Sun, Jan 31, 2016 at 5:00 PM, Jiffin Tony Thottan
>>>>>>>> <jthottan at redhat.com> wrote:
>>>>>>>>>
>>>>>>>>> On 31/01/16 16:19, Serkan Çoban wrote:
>>>>>>>>>> Hi,
>>>>>>>>>> I am testing nfs-ganesha with pNFS on EC volume and I want to ask
>>>>>>>>>> some
>>>>>>>>>> questions.
>>>>>>>>>> Assume we have two clients: c1,c2
>>>>>>>>>> and six servers with one 4+2 EC volume constructed as below:
>>>>>>>>>>
>>>>>>>>>> gluster volume create vol1 disperse 6 redundancy 2
>>>>>>>>>> server{1..6}:/brick/b1
>>>>>>>>>> \
>>>>>>>>>>
>>>>>>>>>>           server{1..6}:/brick/b2 \
>>>>>>>>>>
>>>>>>>>>>           server{1..6}:/brick/b3 \
>>>>>>>>>>
>>>>>>>>>>           server{1..6}:/brick/b4 \
>>>>>>>>>>
>>>>>>>>>>           server{1..6}:/brick/b5 \
>>>>>>>>>>
>>>>>>>>>>           server{1..6}:/brick/b6
>>>>>>>>>> vol1 is mounted on both clients as server1:/vol1
>>>>>>>>>>
>>>>>>>>>> Here is first question: When I write file1 from client1 and file2
>>>>>>>>>> from
>>>>>>>>>> client2; which servers get the files? In my opinion server1 gets
>>>>>>>>>> file1
>>>>>>>>>> and server2 gets file2 and do EC calculations and distribute chunks
>>>>>>>>>> to
>>>>>>>>>> other servers. Am I right?
>>>>>>>>>>
>>>>>>>>>> Can anyone explain detailed read/write path with pNFS and EC
>>>>>>>>>> volumes?
>>>>>>>>>
>>>>>> Currently in pNFS cluster , request is send to the first DS available
>>>>>> in
>>>>>> the
>>>>>> list.
>>>>>> So I am to planning to distribute different files among the available
>>>>>> DS,
>>>>>> as
>>>>>> a first step
>>>>>> I had send out patch in gluster (http://review.gluster.org/#/c/13402/)
>>>>>> After this one got merged, there are certain changes in ganesha side
>>>>>> too.
>>>>>> I had tested both changes in my setup and it was working as accepted.
>>>>>>
>>>>>> Once again thanks for pointing out this issue
>>>>>>
>>>>>> --
>>>>>> Regards,
>>>>>> Jiffin
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>>>> I never tried pNFS with EC volume, will try the same by my own and
>>>>>>>>> reply
>>>>>>>>> to
>>>>>>>>> your question as soon as possible.
>>>>>>>>> --
>>>>>>>>> Jiffin
>>>>>>>>>
>>>>>>>>>> Thanks,
>>>>>>>>>> Serkan
>>>>>>>>>> _______________________________________________
>>>>>>>>>> Gluster-users mailing list
>>>>>>>>>> Gluster-users at gluster.org
>>>>>>>>>> http://www.gluster.org/mailman/listinfo/gluster-users
>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> Gluster-users mailing list
>>>>>>>>> Gluster-users at gluster.org
>>>>>>>>> http://www.gluster.org/mailman/listinfo/gluster-users
>>>>>>>> _______________________________________________
>>>>>>>> Gluster-users mailing list
>>>>>>>> Gluster-users at gluster.org
>>>>>>>> http://www.gluster.org/mailman/listinfo/gluster-users
>>>>>>>