release plan for mdadm

release plan for mdadm

[Tkaczyk, Mariusz]

Hi Jes,

It's been a while since last mdadm release. Mdadm-4.2 release that was
mentioned back in July does not happened yet. It's getting messy to
manage mdadm across multiple distributions.

Also, not all OSVs are willing to cherry-pick the patches, especially
for stable project - like mdadm, so only critical bugfixes are landing
in the distros.
As a result - new OSes has various forks of mdadm-4.1 and the difference
is growing with every backported patch. It leads us to situation where
those forks may have own bugs, caused by many missing bugfixes or wrongly
resolved merge conflicts.
To be honest - it becomes more and more problematic for us to track all
fixes in different supported distros.

We are searching for solutions for those problems and we are counting on
your support:
Short term - is there any way that we can help you to release next version
of mdadm soon?

Long term - what do you think about smaller, more frequent releases of
mdadm? Maybe twice a year is an option (similar to RedHat/Ubuntu
schedule)? That would be better for us and for vendors. They will need
to follow upstream instead resolving bugs reported by us or community.

The benefits will be gained by everyone. User will get up-to-date
software much faster, with minimal vendor input and modifications.
Mdadm bugs will be predictable across distros. We could help with
testing IMSM and basic functionality of native metadata.

What are your thoughts?
Regards,
Mariusz

Re: release plan for mdadm

[Tkaczyk, Mariusz]

On 27.01.2021 12:39, Tkaczyk, Mariusz wrote:
> Hi Jes,
> 
> It's been a while since last mdadm release. Mdadm-4.2 release that was
> mentioned back in July does not happened yet. It's getting messy to
> manage mdadm across multiple distributions.
> 
> Also, not all OSVs are willing to cherry-pick the patches, especially
> for stable project - like mdadm, so only critical bugfixes are landing
> in the distros.
> As a result - new OSes has various forks of mdadm-4.1 and the difference
> is growing with every backported patch. It leads us to situation where
> those forks may have own bugs, caused by many missing bugfixes or wrongly
> resolved merge conflicts.
> To be honest - it becomes more and more problematic for us to track all
> fixes in different supported distros.
> 
> We are searching for solutions for those problems and we are counting on
> your support:
> Short term - is there any way that we can help you to release next version
> of mdadm soon?
> 
> Long term - what do you think about smaller, more frequent releases of
> mdadm? Maybe twice a year is an option (similar to RedHat/Ubuntu
> schedule)? That would be better for us and for vendors. They will need
> to follow upstream instead resolving bugs reported by us or community.
> 
> The benefits will be gained by everyone. User will get up-to-date
> software much faster, with minimal vendor input and modifications.
> Mdadm bugs will be predictable across distros. We could help with
> testing IMSM and basic functionality of native metadata.
> 
> What are your thoughts?
> Regards,
> Mariusz

Hi Jes,
Any feedback?

Mariusz

Re: release plan for mdadm

[Jes Sorensen]

On 1/27/21 6:39 AM, Tkaczyk, Mariusz wrote:
> Hi Jes,
> 
> It's been a while since last mdadm release. Mdadm-4.2 release that was
> mentioned back in July does not happened yet. It's getting messy to
> manage mdadm across multiple distributions.
> 
> Also, not all OSVs are willing to cherry-pick the patches, especially
> for stable project - like mdadm, so only critical bugfixes are landing
> in the distros.
> As a result - new OSes has various forks of mdadm-4.1 and the difference
> is growing with every backported patch. It leads us to situation where
> those forks may have own bugs, caused by many missing bugfixes or wrongly
> resolved merge conflicts.
> To be honest - it becomes more and more problematic for us to track all
> fixes in different supported distros.
> 
> We are searching for solutions for those problems and we are counting on
> your support:
> Short term - is there any way that we can help you to release next version
> of mdadm soon?
> 
> Long term - what do you think about smaller, more frequent releases of
> mdadm? Maybe twice a year is an option (similar to RedHat/Ubuntu
> schedule)? That would be better for us and for vendors. They will need
> to follow upstream instead resolving bugs reported by us or community.
> 
> The benefits will be gained by everyone. User will get up-to-date
> software much faster, with minimal vendor input and modifications.
> Mdadm bugs will be predictable across distros. We could help with
> testing IMSM and basic functionality of native metadata.

Hi Mariusz,

Sorry for the slow response. Our daughter was born in late December and
I was on paternity leave through Feb 5, so still catching up. I also
switched teams at work back in July so my focus was shifted.

I'd very much like to see a release, and we should do one quick. Doing
more regular releases will also make it easier to ship them, so I am not
against that at all.

I am not aware of anything major pending right now, so if we can get
focus on any pending patches and get them in over the next week or two,
then I can cut an -rc and we can do a release soon. Especially if you
can help out regression testing the -rc candidate(s).

Cheers,
Jes

Re: release plan for mdadm

[Tkaczyk, Mariusz]

On 02.03.2021 23:50, Jes Sorensen wrote:
> On 1/27/21 6:39 AM, Tkaczyk, Mariusz wrote:
>> Hi Jes,
>>
>> It's been a while since last mdadm release. Mdadm-4.2 release that was
>> mentioned back in July does not happened yet. It's getting messy to
>> manage mdadm across multiple distributions.
>>
>> Also, not all OSVs are willing to cherry-pick the patches, especially
>> for stable project - like mdadm, so only critical bugfixes are landing
>> in the distros.
>> As a result - new OSes has various forks of mdadm-4.1 and the difference
>> is growing with every backported patch. It leads us to situation where
>> those forks may have own bugs, caused by many missing bugfixes or wrongly
>> resolved merge conflicts.
>> To be honest - it becomes more and more problematic for us to track all
>> fixes in different supported distros.
>>
>> We are searching for solutions for those problems and we are counting on
>> your support:
>> Short term - is there any way that we can help you to release next version
>> of mdadm soon?
>>
>> Long term - what do you think about smaller, more frequent releases of
>> mdadm? Maybe twice a year is an option (similar to RedHat/Ubuntu
>> schedule)? That would be better for us and for vendors. They will need
>> to follow upstream instead resolving bugs reported by us or community.
>>
>> The benefits will be gained by everyone. User will get up-to-date
>> software much faster, with minimal vendor input and modifications.
>> Mdadm bugs will be predictable across distros. We could help with
>> testing IMSM and basic functionality of native metadata.
> 
> Hi Mariusz,
> 
> Sorry for the slow response. Our daughter was born in late December and
> I was on paternity leave through Feb 5, so still catching up. I also
> switched teams at work back in July so my focus was shifted.
Hi Jes,
Congratulations:)
> 
> I'd very much like to see a release, and we should do one quick. Doing
> more regular releases will also make it easier to ship them, so I am not
> against that at all.
> 
> I am not aware of anything major pending right now, so if we can get
> focus on any pending patches and get them in over the next week or two,
> then I can cut an -rc and we can do a release soon. Especially if you
> can help out regression testing the -rc candidate(s).
> 
> Cheers,
> Jes
> 
Thanks for answer. Please review all patches in queue, and mark -rc.
Then I will schedule regression.

Mariusz

Re: release plan for mdadm

[Piergiorgio Sartor]

On Tue, Mar 02, 2021 at 05:50:10PM -0500, Jes Sorensen wrote:
> Sorry for the slow response. Our daughter was born in late December and

Well! Congratulations!

I know, a bit OT, but still deserved!

bye,

-- 

piergiorgio

Re: release plan for mdadm

[Radtke, Jakub]

Hi Jes,
     I have also resent the patches related to the bitmap support in 
IMSM posted some time ago.
Regards, Jakub

> On 02.03.2021 23:50, Jes Sorensen wrote:
>> On 1/27/21 6:39 AM, Tkaczyk, Mariusz wrote:
>>> Hi Jes,
>>>
>>> It's been a while since last mdadm release. Mdadm-4.2 release that was
>>> mentioned back in July does not happened yet. It's getting messy to
>>> manage mdadm across multiple distributions.
>>>
>>> Also, not all OSVs are willing to cherry-pick the patches, especially
>>> for stable project - like mdadm, so only critical bugfixes are landing
>>> in the distros.
>>> As a result - new OSes has various forks of mdadm-4.1 and the 
>>> difference
>>> is growing with every backported patch. It leads us to situation where
>>> those forks may have own bugs, caused by many missing bugfixes or 
>>> wrongly
>>> resolved merge conflicts.
>>> To be honest - it becomes more and more problematic for us to track all
>>> fixes in different supported distros.
>>>
>>> We are searching for solutions for those problems and we are 
>>> counting on
>>> your support:
>>> Short term - is there any way that we can help you to release next 
>>> version
>>> of mdadm soon?
>>>
>>> Long term - what do you think about smaller, more frequent releases of
>>> mdadm? Maybe twice a year is an option (similar to RedHat/Ubuntu
>>> schedule)? That would be better for us and for vendors. They will need
>>> to follow upstream instead resolving bugs reported by us or community.
>>>
>>> The benefits will be gained by everyone. User will get up-to-date
>>> software much faster, with minimal vendor input and modifications.
>>> Mdadm bugs will be predictable across distros. We could help with
>>> testing IMSM and basic functionality of native metadata.
>>
>> Hi Mariusz,
>>
>> Sorry for the slow response. Our daughter was born in late December and
>> I was on paternity leave through Feb 5, so still catching up. I also
>> switched teams at work back in July so my focus was shifted.
> Hi Jes,
> Congratulations:)
>>
>> I'd very much like to see a release, and we should do one quick. Doing
>> more regular releases will also make it easier to ship them, so I am not
>> against that at all.
>>
>> I am not aware of anything major pending right now, so if we can get
>> focus on any pending patches and get them in over the next week or two,
>> then I can cut an -rc and we can do a release soon. Especially if you
>> can help out regression testing the -rc candidate(s).
>>
>> Cheers,
>> Jes
>>
> Thanks for answer. Please review all patches in queue, and mark -rc.
> Then I will schedule regression.
>
> Mariusz

Re: release plan for mdadm

[Jes Sorensen]

On 3/3/21 3:13 AM, Tkaczyk, Mariusz wrote:
> On 02.03.2021 23:50, Jes Sorensen wrote:
>> I am not aware of anything major pending right now, so if we can get
>> focus on any pending patches and get them in over the next week or two,
>> then I can cut an -rc and we can do a release soon. Especially if you
>> can help out regression testing the -rc candidate(s).
>>
>> Cheers,
>> Jes
>>
> Thanks for answer. Please review all patches in queue, and mark -rc.
> Then I will schedule regression.

I have applied everything in my pipeline, except for Oleksandr's

[PATCH] imsm: nvme multipath support

Once we have that one resolved, I can tag -rc. Which also makes this a
last call for missing patches before the -rc.

Cheers,
Jes

Re: release plan for mdadm

[Felix Lechner]

Hi Jes,

Congratulations to the birth of your daughter!

On Tue, Mar 2, 2021 at 7:38 PM Jes Sorensen <jes@trained-monkey.org> wrote:
>
> I am not aware of anything major pending

Please allow me to mention a fresh, critical Debian bug [1] with data
loss when /proc, /sys and /dev are not mounted.

I assumed maintenance recently, and the Debian package lags behind
your development status. Perhaps this issue was already addressed.
Thanks!

Kind regards
Felix Lechner

[1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=982459