Apply against 2.6.34-rc3 pnfs branch. For pNFS client submission. Only
touches
the block layout driver.
This patch set implements support for the complex volumes structures such
as
stripe of multiple volumes in a single metavolume and slice of metavolumes
into
smaller metavolumes. The patch set is specific to pnfs blocklayout and uses
simple rpc pipefs to generate messages that are passed to the userland
which is
using an daemon to discover all the devices in the message. The userspace
daemon
patches will be sent in a separate patch set.
The patch set replaces the current single volume implementation of the
pnfs blocklayout
that has performance limitations and cannot support wide striped file
systems.
The wide stripe is imperious to the performance scalability both for
single and multiple
clients configurations. These patches address only the pnfs blocklayout
client. Similar
patches will be implemented for the pnfsd block layout served in the
second phase.
0001-pnfs-block-layout-add-simple-rpc-pipefs.patch
0002-pnfs-block-layout-remove-device-creation-or-remove-from-kernel.patch
0003-pnfs-block-layout-add-block-device-pipe-processing-based-on-simple-rpc.patch
Testing:
CONFIG_NFS_V4_1 set:
NFSv4.1/pNFS mounts:
Connectathon tests pass against Celerra/pNFS block layout server with both
simple
volumes and complex volumes stripe on 4 volumes and using a slice of half
of the
striped metavolume.
CONFIG_NFS_V4_1 not set:
NFSv4.0 mount passes Connectathon tests.
/Sorin
On Thu, 10 Jun 2010 11:44:09 -0400, Benny Halevy <[email protected]>
wrote:
> On Jun. 04, 2010, 21:09 +0300, sfaibish <[email protected]> wrote:
>> Apply against 2.6.34-rc3 pnfs branch. For pNFS client submission. Only
>> touches
>> the block layout driver.
>
> Sorin, I suggest you guys test this set with the linux-pnfs 2.6.35-rc2
> tree
We already tested the patches and they work in our QA lab before we sent
the
patches. We also tested in our labs in France. We used F13 as the base.
> against both the EMC and the spnfs-block server in the Ann Arbor
As I mentioned in the patch at this time the spnfs block server doesn't
support complex structures. After we get the pnfs client patches in the
git we will work to implement the complex volumes support in the spnfs
server. In principle the testing was done with EMC pnfs server and all
the cthon tests passed so this is what we would do at BAT anyways.
I have VM with the code so, if Bruce can have a VM client with the
patches we have the pnfs server of EMC in CITI lab.
> Bakeathon
> and resend.
Unfortunatelly, I will not be there and I am not sure if the new pNFS block
developer from CITI will be up to speed. I am busy with the next BAT
organization.
>
> I admit I didn't review this thoroughly yet and hopefully we can find
> some
> time to do it together in Ann Arbor.
:(
>
> Benny
>
>>
>> This patch set implements support for the complex volumes structures
>> such
>> as
>> stripe of multiple volumes in a single metavolume and slice of
>> metavolumes
>> into
>> smaller metavolumes. The patch set is specific to pnfs blocklayout and
>> uses
>> simple rpc pipefs to generate messages that are passed to the userland
>> which is
>> using an daemon to discover all the devices in the message. The
>> userspace
>> daemon
>> patches will be sent in a separate patch set.
>>
>> The patch set replaces the current single volume implementation of the
>> pnfs blocklayout
>> that has performance limitations and cannot support wide striped file
>> systems.
>> The wide stripe is imperious to the performance scalability both for
>> single and multiple
>> clients configurations. These patches address only the pnfs blocklayout
>> client. Similar
>> patches will be implemented for the pnfsd block layout served in the
>> second phase.
>>
>> 0001-pnfs-block-layout-add-simple-rpc-pipefs.patch
>> 0002-pnfs-block-layout-remove-device-creation-or-remove-from-kernel.patch
>> 0003-pnfs-block-layout-add-block-device-pipe-processing-based-on-simple-rpc.patch
>>
>> Testing:
>> CONFIG_NFS_V4_1 set:
>>
>> NFSv4.1/pNFS mounts:
>> Connectathon tests pass against Celerra/pNFS block layout server with
>> both
>> simple
>> volumes and complex volumes stripe on 4 volumes and using a slice of
>> half
>> of the
>> striped metavolume.
>>
>> CONFIG_NFS_V4_1 not set:
>> NFSv4.0 mount passes Connectathon tests.
>>
>> /Sorin
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
>> the body of a message to [email protected]
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
>
--
Best Regards
Sorin Faibish
Corporate Distinguished Engineer
Network Storage Group
EMC?
where information lives
Phone: 508-435-1000 x 48545
Cellphone: 617-510-0422
Email : [email protected]
On Jun. 04, 2010, 21:09 +0300, sfaibish <[email protected]> wrote:
> Apply against 2.6.34-rc3 pnfs branch. For pNFS client submission. Only
> touches
> the block layout driver.
Sorin, I suggest you guys test this set with the linux-pnfs 2.6.35-rc2 tree
against both the EMC and the spnfs-block server in the Ann Arbor Bakeathon
and resend.
I admit I didn't review this thoroughly yet and hopefully we can find some
time to do it together in Ann Arbor.
Benny
>
> This patch set implements support for the complex volumes structures such
> as
> stripe of multiple volumes in a single metavolume and slice of metavolumes
> into
> smaller metavolumes. The patch set is specific to pnfs blocklayout and uses
> simple rpc pipefs to generate messages that are passed to the userland
> which is
> using an daemon to discover all the devices in the message. The userspace
> daemon
> patches will be sent in a separate patch set.
>
> The patch set replaces the current single volume implementation of the
> pnfs blocklayout
> that has performance limitations and cannot support wide striped file
> systems.
> The wide stripe is imperious to the performance scalability both for
> single and multiple
> clients configurations. These patches address only the pnfs blocklayout
> client. Similar
> patches will be implemented for the pnfsd block layout served in the
> second phase.
>
> 0001-pnfs-block-layout-add-simple-rpc-pipefs.patch
> 0002-pnfs-block-layout-remove-device-creation-or-remove-from-kernel.patch
> 0003-pnfs-block-layout-add-block-device-pipe-processing-based-on-simple-rpc.patch
>
> Testing:
> CONFIG_NFS_V4_1 set:
>
> NFSv4.1/pNFS mounts:
> Connectathon tests pass against Celerra/pNFS block layout server with both
> simple
> volumes and complex volumes stripe on 4 volumes and using a slice of half
> of the
> striped metavolume.
>
> CONFIG_NFS_V4_1 not set:
> NFSv4.0 mount passes Connectathon tests.
>
> /Sorin
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
On Jun. 10, 2010, 20:07 +0300, sfaibish <[email protected]> wrote:
> On Thu, 10 Jun 2010 11:44:09 -0400, Benny Halevy <[email protected]>
> wrote:
>
>> On Jun. 04, 2010, 21:09 +0300, sfaibish <[email protected]> wrote:
>>> Apply against 2.6.34-rc3 pnfs branch. For pNFS client submission. Only
>>> touches
>>> the block layout driver.
>>
>> Sorin, I suggest you guys test this set with the linux-pnfs 2.6.35-rc2
>> tree
> We already tested the patches and they work in our QA lab before we sent
> the
> patches. We also tested in our labs in France. We used F13 as the base.
>
>> against both the EMC and the spnfs-block server in the Ann Arbor
> As I mentioned in the patch at this time the spnfs block server doesn't
> support complex structures. After we get the pnfs client patches in the
> git we will work to implement the complex volumes support in the spnfs
> server. In principle the testing was done with EMC pnfs server and all
> the cthon tests passed so this is what we would do at BAT anyways.
> I have VM with the code so, if Bruce can have a VM client with the
> patches we have the pnfs server of EMC in CITI lab.
>
OK, but I want to sure the new code is backward compatible with
the spnfs-block server, as it is today.
Benny
>> Bakeathon
>> and resend.
> Unfortunatelly, I will not be there and I am not sure if the new pNFS block
> developer from CITI will be up to speed. I am busy with the next BAT
> organization.
>
>>
>> I admit I didn't review this thoroughly yet and hopefully we can find
>> some
>> time to do it together in Ann Arbor.
> :(
>
>>
>> Benny
>>
>>>
>>> This patch set implements support for the complex volumes structures
>>> such
>>> as
>>> stripe of multiple volumes in a single metavolume and slice of
>>> metavolumes
>>> into
>>> smaller metavolumes. The patch set is specific to pnfs blocklayout and
>>> uses
>>> simple rpc pipefs to generate messages that are passed to the userland
>>> which is
>>> using an daemon to discover all the devices in the message. The
>>> userspace
>>> daemon
>>> patches will be sent in a separate patch set.
>>>
>>> The patch set replaces the current single volume implementation of the
>>> pnfs blocklayout
>>> that has performance limitations and cannot support wide striped file
>>> systems.
>>> The wide stripe is imperious to the performance scalability both for
>>> single and multiple
>>> clients configurations. These patches address only the pnfs blocklayout
>>> client. Similar
>>> patches will be implemented for the pnfsd block layout served in the
>>> second phase.
>>>
>>> 0001-pnfs-block-layout-add-simple-rpc-pipefs.patch
>>> 0002-pnfs-block-layout-remove-device-creation-or-remove-from-kernel.patch
>>> 0003-pnfs-block-layout-add-block-device-pipe-processing-based-on-simple-rpc.patch
>>>
>>> Testing:
>>> CONFIG_NFS_V4_1 set:
>>>
>>> NFSv4.1/pNFS mounts:
>>> Connectathon tests pass against Celerra/pNFS block layout server with
>>> both
>>> simple
>>> volumes and complex volumes stripe on 4 volumes and using a slice of
>>> half
>>> of the
>>> striped metavolume.
>>>
>>> CONFIG_NFS_V4_1 not set:
>>> NFSv4.0 mount passes Connectathon tests.
>>>
>>> /Sorin
>>>
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
>>> the body of a message to [email protected]
>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
>> the body of a message to [email protected]
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
>>
>
>
>
On Thu, 10 Jun 2010 13:10:27 -0400, Benny Halevy <[email protected]> =
=20
wrote:
> On Jun. 10, 2010, 20:07 +0300, sfaibish <[email protected]> wrote:
>> On Thu, 10 Jun 2010 11:44:09 -0400, Benny Halevy <bhalevy-C4P08NqkoRmIwRZHo2/[email protected]=
m>
>> wrote:
>>
>>> On Jun. 04, 2010, 21:09 +0300, sfaibish <[email protected]> wrote:
>>>> Apply against 2.6.34-rc3 pnfs branch. For pNFS client submission. =
Only
>>>> touches
>>>> the block layout driver.
>>>
>>> Sorin, I suggest you guys test this set with the linux-pnfs 2.6.35-=
rc2
>>> tree
>> We already tested the patches and they work in our QA lab before we =
sent
>> the
>> patches. We also tested in our labs in France. We used F13 as the ba=
se.
>>
>>> against both the EMC and the spnfs-block server in the Ann Arbor
>> As I mentioned in the patch at this time the spnfs block server does=
n't
>> support complex structures. After we get the pnfs client patches in =
the
>> git we will work to implement the complex volumes support in the spn=
fs
>> server. In principle the testing was done with EMC pnfs server and a=
ll
>> the cthon tests passed so this is what we would do at BAT anyways.
>> I have VM with the code so, if Bruce can have a VM client with the
>> patches we have the pnfs server of EMC in CITI lab.
>>
>
> OK, but I want to sure the new code is backward compatible with
> the spnfs-block server, as it is today.
Yes, that we also tested and we were passing the cthon tests as before.
This is part of our QA test suite. We also tested support of single vol=
ume
=46S.
/Sorin
>
> Benny
>
>>> Bakeathon
>>> and resend.
>> Unfortunatelly, I will not be there and I am not sure if the new pNF=
S =20
>> block
>> developer from CITI will be up to speed. I am busy with the next BAT
>> organization.
>>
>>>
>>> I admit I didn't review this thoroughly yet and hopefully we can fi=
nd
>>> some
>>> time to do it together in Ann Arbor.
>> :(
>>
>>>
>>> Benny
>>>
>>>>
>>>> This patch set implements support for the complex volumes structur=
es
>>>> such
>>>> as
>>>> stripe of multiple volumes in a single metavolume and slice of
>>>> metavolumes
>>>> into
>>>> smaller metavolumes. The patch set is specific to pnfs blocklayout=
and
>>>> uses
>>>> simple rpc pipefs to generate messages that are passed to the user=
land
>>>> which is
>>>> using an daemon to discover all the devices in the message. The
>>>> userspace
>>>> daemon
>>>> patches will be sent in a separate patch set.
>>>>
>>>> The patch set replaces the current single volume implementation of=
the
>>>> pnfs blocklayout
>>>> that has performance limitations and cannot support wide striped f=
ile
>>>> systems.
>>>> The wide stripe is imperious to the performance scalability both f=
or
>>>> single and multiple
>>>> clients configurations. These patches address only the pnfs =20
>>>> blocklayout
>>>> client. Similar
>>>> patches will be implemented for the pnfsd block layout served in t=
he
>>>> second phase.
>>>>
>>>> 0001-pnfs-block-layout-add-simple-rpc-pipefs.patch
>>>> 0002-pnfs-block-layout-remove-device-creation-or-remove-from-kerne=
l.patch
>>>> 0003-pnfs-block-layout-add-block-device-pipe-processing-based-on-s=
imple-rpc.patch
>>>>
>>>> Testing:
>>>> CONFIG_NFS_V4_1 set:
>>>>
>>>> NFSv4.1/pNFS mounts:
>>>> Connectathon tests pass against Celerra/pNFS block layout server w=
ith
>>>> both
>>>> simple
>>>> volumes and complex volumes stripe on 4 volumes and using a slice =
of
>>>> half
>>>> of the
>>>> striped metavolume.
>>>>
>>>> CONFIG_NFS_V4_1 not set:
>>>> NFSv4.0 mount passes Connectathon tests.
>>>>
>>>> /Sorin
>>>>
>>>> --
>>>> To unsubscribe from this list: send the line "unsubscribe linux-nf=
s" =20
>>>> in
>>>> the body of a message to [email protected]
>>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>>>
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-nfs=
" in
>>> the body of a message to [email protected]
>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>>
>>>
>>
>>
>>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" =
in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
>
--=20
Best Regards
Sorin Faibish
Corporate Distinguished Engineer
Network Storage Group
EMC=B2
where information lives
Phone: 508-435-1000 x 48545
Cellphone: 617-510-0422
Email : [email protected]