Re: Ceph disk

Re: Ceph disk

[Loic Dachary]

[-- Attachment #1: Type: text/plain, Size: 2516 bytes --]

[cc'ing ceph-devel public mailing list]

Hi,

On 31/03/2015 20:47, Kewaan Ejaz wrote:
> Hello Loic,
> 
> Thank you again for the opportunity to contribute to ceph. I am not sure
> how am I suppose to proceed but I have the following things in mind;

Thanks for proposing to help :-)

> 1) Fully understanding the ceph-disk code
> 2) I would start using the ceph-disk in my sandbox environment
> (Virtualbox)

You'll find that it's fairly straightforward: one file, all in it. There is room for improvement.

https://ceph.com/git/?p=ceph.git;a=blob;f=src/test/ceph-disk.sh are the tests that make check will run 
https://ceph.com/git/?p=ceph.git;a=blob;f=src/test/ceph-disk-root.sh are the tests that make check will run if ./configure --enable-root-make-check is used but you can also run one manually with cd src ; sudo test/ceph-disk.sh test_activate_dev and it will use the /dev/loop device to simulate a disk. A few sanity checks were added recently to verify that the loop device has been loaded with the necessary loop.max_part=16 (or more) parameter and that /dev/disk/part-byuuid is populated as expected. 

> I think I would require a week to get back to you. But other than that,
> let me know what kind of VM do you have in mind. For example,
> 
> 1) What kind of OS would you prefer from the Openstack cluster?

I feel more confortable with Debian or Ubuntu but CentOS or Fedora are also fine. What is required for test purposes is a dedicated tenant with the ability to run two virtual machines, 1GB RAM, 10GB disk, 1core. We would need a variety of images to test against (Ubuntu 12.04 + 14.04, Debian jessie, CentOS 7 + 6, Fedora 20 + 21, OpenSUSE 13.2).

> 2) We are not using Cinder or Swift in our Openstack cluster. Is that a
> Problem?

That will be fine: the loop device can be used as a spare disk instead of provisioning one with cinder. 

As soon as you can have that OpenStack tenant ready, it can be used with https://github.com/osynge/whatenv to run tests for https://github.com/ceph/ceph/pull/4036 and verify that the detection system actually works as expected. 

And once it's done, we could make it so this test is run whenever a pull request is posted that modifies ceph-disk and ensure we're not risking a regression. It would be a very valuable service. Not too CPU intensive either since we're not seeing more than a few patches on ceph-disk every month.

Cheers

> Best Regards,
> Kewaan.


-- 
Loïc Dachary, Artisan Logiciel Libre


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: Ceph disk

[Kewaan Ejaz]

Hi Folks,

Sorry for the lack of updates from my side. Actually, we had a few crashes in our openstack cluster (nothing out of the norm) but due to easter holidays, manpower has been quite tight. I am also working on recovering/setting up those lost compute nodes to make the resources available for our environment. I will update you all
once the pressure is gone. Thanks for your understanding!

Best Regards,
Kewaan.
On Tue, 2015-03-31 at 22:02 +0200, Loic Dachary wrote:
> [cc'ing ceph-devel public mailing list]
> 
> Hi,
> 
> On 31/03/2015 20:47, Kewaan Ejaz wrote:
> > Hello Loic,
> > 
> > Thank you again for the opportunity to contribute to ceph. I am not sure
> > how am I suppose to proceed but I have the following things in mind;
> 
> Thanks for proposing to help :-)
> 
> > 1) Fully understanding the ceph-disk code
> > 2) I would start using the ceph-disk in my sandbox environment
> > (Virtualbox)
> 
> You'll find that it's fairly straightforward: one file, all in it. There is room for improvement.
> 
> https://ceph.com/git/?p=ceph.git;a=blob;f=src/test/ceph-disk.sh are the tests that make check will run 
> https://ceph.com/git/?p=ceph.git;a=blob;f=src/test/ceph-disk-root.sh are the tests that make check will run if ./configure --enable-root-make-check is used but you can also run one manually with cd src ; sudo test/ceph-disk.sh test_activate_dev and it will use the /dev/loop device to simulate a disk. A few sanity checks were added recently to verify that the loop device has been loaded with the necessary loop.max_part=16 (or more) parameter and that /dev/disk/part-byuuid is populated as expected. 
> 
> > I think I would require a week to get back to you. But other than that,
> > let me know what kind of VM do you have in mind. For example,
> > 
> > 1) What kind of OS would you prefer from the Openstack cluster?
> 
> I feel more confortable with Debian or Ubuntu but CentOS or Fedora are also fine. What is required for test purposes is a dedicated tenant with the ability to run two virtual machines, 1GB RAM, 10GB disk, 1core. We would need a variety of images to test against (Ubuntu 12.04 + 14.04, Debian jessie, CentOS 7 + 6, Fedora 20 + 21, OpenSUSE 13.2).
> 
> > 2) We are not using Cinder or Swift in our Openstack cluster. Is that a
> > Problem?
> 
> That will be fine: the loop device can be used as a spare disk instead of provisioning one with cinder. 
> 
> As soon as you can have that OpenStack tenant ready, it can be used with https://github.com/osynge/whatenv to run tests for https://github.com/ceph/ceph/pull/4036 and verify that the detection system actually works as expected. 
> 
> And once it's done, we could make it so this test is run whenever a pull request is posted that modifies ceph-disk and ensure we're not risking a regression. It would be a very valuable service. Not too CPU intensive either since we're not seeing more than a few patches on ceph-disk every month.
> 
> Cheers
> 
> > Best Regards,
> > Kewaan.
> 
> 

Re: Ceph disk

[Loic Dachary]

[-- Attachment #1: Type: text/plain, Size: 3203 bytes --]



On 07/04/2015 15:17, Kewaan Ejaz wrote:
> 
> Hi Folks,
> 
> Sorry for the lack of updates from my side. Actually, we had a few crashes in our openstack cluster (nothing out of the norm) but due to easter holidays, manpower has been quite tight. I am also working on recovering/setting up those lost compute nodes to make the resources available for our environment. I will update you all
> once the pressure is gone. Thanks for your understanding!

Thanks for the update !

> 
> Best Regards,
> Kewaan.
> On Tue, 2015-03-31 at 22:02 +0200, Loic Dachary wrote:
>> [cc'ing ceph-devel public mailing list]
>>
>> Hi,
>>
>> On 31/03/2015 20:47, Kewaan Ejaz wrote:
>>> Hello Loic,
>>>
>>> Thank you again for the opportunity to contribute to ceph. I am not sure
>>> how am I suppose to proceed but I have the following things in mind;
>>
>> Thanks for proposing to help :-)
>>
>>> 1) Fully understanding the ceph-disk code
>>> 2) I would start using the ceph-disk in my sandbox environment
>>> (Virtualbox)
>>
>> You'll find that it's fairly straightforward: one file, all in it. There is room for improvement.
>>
>> https://ceph.com/git/?p=ceph.git;a=blob;f=src/test/ceph-disk.sh are the tests that make check will run 
>> https://ceph.com/git/?p=ceph.git;a=blob;f=src/test/ceph-disk-root.sh are the tests that make check will run if ./configure --enable-root-make-check is used but you can also run one manually with cd src ; sudo test/ceph-disk.sh test_activate_dev and it will use the /dev/loop device to simulate a disk. A few sanity checks were added recently to verify that the loop device has been loaded with the necessary loop.max_part=16 (or more) parameter and that /dev/disk/part-byuuid is populated as expected. 
>>
>>> I think I would require a week to get back to you. But other than that,
>>> let me know what kind of VM do you have in mind. For example,
>>>
>>> 1) What kind of OS would you prefer from the Openstack cluster?
>>
>> I feel more confortable with Debian or Ubuntu but CentOS or Fedora are also fine. What is required for test purposes is a dedicated tenant with the ability to run two virtual machines, 1GB RAM, 10GB disk, 1core. We would need a variety of images to test against (Ubuntu 12.04 + 14.04, Debian jessie, CentOS 7 + 6, Fedora 20 + 21, OpenSUSE 13.2).
>>
>>> 2) We are not using Cinder or Swift in our Openstack cluster. Is that a
>>> Problem?
>>
>> That will be fine: the loop device can be used as a spare disk instead of provisioning one with cinder. 
>>
>> As soon as you can have that OpenStack tenant ready, it can be used with https://github.com/osynge/whatenv to run tests for https://github.com/ceph/ceph/pull/4036 and verify that the detection system actually works as expected. 
>>
>> And once it's done, we could make it so this test is run whenever a pull request is posted that modifies ceph-disk and ensure we're not risking a regression. It would be a very valuable service. Not too CPU intensive either since we're not seeing more than a few patches on ceph-disk every month.
>>
>> Cheers
>>
>>> Best Regards,
>>> Kewaan.
>>
>>
> 
> 
> 

-- 
Loïc Dachary, Artisan Logiciel Libre


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]