mdadm regression tests fail

mdadm regression tests fail

[Bruce Dubbs]

I am trying to document the mdadm-4.2 installation procedures for our book,
https://www.linuxfromscratch.org/blfs/view/svn/postlfs/mdadm.html

For testing, I am doing a simple:

    make
    sudo ./test --keep-going --logdir=test-logs --save-logs

But I get failures for about half the tests.

Digging in a bit I just ran:

   sudo ./test --tests=00raid0 --logdir=test-logs

This is the first test that fails.  With some hacking, it appears that the first
portion of this test that fails is:

     mdadm -CR $md0 -e0.90 -l0 -n4 $dev0 $dev1 $dev2 $dev3

This resolves to

     mdadm -CR /dev/md0 -e0.90 -l0 -n4 /dev/loop0 /dev/loop1 /dev/loop2 /dev/loop3

There is not a lot of error output in the test, so I manually ran:

     dd if=/dev/zero of=/tmp/mdtest0 count=20000 bs=1K
     losetup /dev/loop0 /tmp/mdtest0

For /dev/loop[0123]

Then I ran

     mdadm -CR /dev/md0 -e0.90 -l0 -n4 /dev/loop0 /dev/loop1 /dev/loop2 /dev/loop3
     mdadm: 0.90 metadata does not support layouts for RAID0

My question is whether the regression tests in the tarball are valid for mdadm-4.2?

     -- Bruce Dubbs
        linuxfromscratch.org

Note: The kernel is version 5.15.12.

Re: mdadm regression tests fail

[Randy Dunlap]

Hi.
[adding linux-raid mailing list]


On 1/4/22 10:55, Bruce Dubbs wrote:
> I am trying to document the mdadm-4.2 installation procedures for our book,
> https://www.linuxfromscratch.org/blfs/view/svn/postlfs/mdadm.html
> 
> For testing, I am doing a simple:
> 
>   make
>   sudo ./test --keep-going --logdir=test-logs --save-logs
> 
> But I get failures for about half the tests.
> 
> Digging in a bit I just ran:
> 
>  sudo ./test --tests=00raid0 --logdir=test-logs
> 
> This is the first test that fails.  With some hacking, it appears that the first portion of this test that fails is:
> 
>   mdadm -CR $md0 -e0.90 -l0 -n4 $dev0 $dev1 $dev2 $dev3
> 
> This resolves to
> 
>   mdadm -CR /dev/md0 -e0.90 -l0 -n4 /dev/loop0 /dev/loop1 /dev/loop2 /dev/loop3
> 
> There is not a lot of error output in the test, so I manually ran:
> 
>   dd if=/dev/zero of=/tmp/mdtest0 count=20000 bs=1K
>   losetup /dev/loop0 /tmp/mdtest0
> 
> For /dev/loop[0123]
> 
> Then I ran
> 
>   mdadm -CR /dev/md0 -e0.90 -l0 -n4 /dev/loop0 /dev/loop1 /dev/loop2 /dev/loop3
>   mdadm: 0.90 metadata does not support layouts for RAID0
> 
> My question is whether the regression tests in the tarball are valid for mdadm-4.2?
> 
>   -- Bruce Dubbs
>      linuxfromscratch.org
> 
> Note: The kernel is version 5.15.12.

-- 
~Randy

Re: mdadm regression tests fail

[Wols Lists]

Bear in mind raid superblock v0.9 is deprecated as in "if it breaks it 
won't be fixed for you".

So I would skip this test, and if you're mentioning raid in the 
handbook, tell people they need to use one of the v1.x formats.

(NB - you can always point them at the linux raid wiki.)

Cheers,
Wol

On 05/01/2022 17:12, Randy Dunlap wrote:
> Hi.
> [adding linux-raid mailing list]
> 
> 
> On 1/4/22 10:55, Bruce Dubbs wrote:
>> I am trying to document the mdadm-4.2 installation procedures for our book,
>> https://www.linuxfromscratch.org/blfs/view/svn/postlfs/mdadm.html
>>
>> For testing, I am doing a simple:
>>
>>    make
>>    sudo ./test --keep-going --logdir=test-logs --save-logs
>>
>> But I get failures for about half the tests.
>>
>> Digging in a bit I just ran:
>>
>>   sudo ./test --tests=00raid0 --logdir=test-logs
>>
>> This is the first test that fails.  With some hacking, it appears that the first portion of this test that fails is:
>>
>>    mdadm -CR $md0 -e0.90 -l0 -n4 $dev0 $dev1 $dev2 $dev3
>>
>> This resolves to
>>
>>    mdadm -CR /dev/md0 -e0.90 -l0 -n4 /dev/loop0 /dev/loop1 /dev/loop2 /dev/loop3
>>
>> There is not a lot of error output in the test, so I manually ran:
>>
>>    dd if=/dev/zero of=/tmp/mdtest0 count=20000 bs=1K
>>    losetup /dev/loop0 /tmp/mdtest0
>>
>> For /dev/loop[0123]
>>
>> Then I ran
>>
>>    mdadm -CR /dev/md0 -e0.90 -l0 -n4 /dev/loop0 /dev/loop1 /dev/loop2 /dev/loop3
>>    mdadm: 0.90 metadata does not support layouts for RAID0
>>
>> My question is whether the regression tests in the tarball are valid for mdadm-4.2?
>>
>>    -- Bruce Dubbs
>>       linuxfromscratch.org
>>
>> Note: The kernel is version 5.15.12.
> 

Re: mdadm regression tests fail

[Bruce Dubbs]

My point is that many of the tests fail.  It's not that someone should use the 
superblock v0.9.  That's only an example. The test should be removed or marked 
"Expected FAIL" or similar.  Our users run the tests as a confidence check that the 
build is successful.  They are generally not trying to debug the package.

I can certainly say that the tests are broken and leave it at that.  If it were only 
a couple of tests that fail, we generally say something like testA and testG are 
known to fail, but in this case fully half of the tests fail.

I would like to know what the maintainers think of the regression tests.  Are they 
maintained?  Should they all pass?  For our users there are far too many tests to run 
them individually.

   -- Bruce


On 1/5/22 11:44 AM, Wols Lists wrote:
> Bear in mind raid superblock v0.9 is deprecated as in "if it breaks it won't be fixed 
> for you".
> 
> So I would skip this test, and if you're mentioning raid in the handbook, tell people 
> they need to use one of the v1.x formats.
> 
> (NB - you can always point them at the linux raid wiki.)
> 
> Cheers,
> Wol
> 
> On 05/01/2022 17:12, Randy Dunlap wrote:
>> Hi.
>> [adding linux-raid mailing list]
>>
>>
>> On 1/4/22 10:55, Bruce Dubbs wrote:
>>> I am trying to document the mdadm-4.2 installation procedures for our book,
>>> https://www.linuxfromscratch.org/blfs/view/svn/postlfs/mdadm.html
>>>
>>> For testing, I am doing a simple:
>>>
>>>    make
>>>    sudo ./test --keep-going --logdir=test-logs --save-logs
>>>
>>> But I get failures for about half the tests.
>>>
>>> Digging in a bit I just ran:
>>>
>>>   sudo ./test --tests=00raid0 --logdir=test-logs
>>>
>>> This is the first test that fails.  With some hacking, it appears that the first 
>>> portion of this test that fails is:
>>>
>>>    mdadm -CR $md0 -e0.90 -l0 -n4 $dev0 $dev1 $dev2 $dev3
>>>
>>> This resolves to
>>>
>>>    mdadm -CR /dev/md0 -e0.90 -l0 -n4 /dev/loop0 /dev/loop1 /dev/loop2 /dev/loop3
>>>
>>> There is not a lot of error output in the test, so I manually ran:
>>>
>>>    dd if=/dev/zero of=/tmp/mdtest0 count=20000 bs=1K
>>>    losetup /dev/loop0 /tmp/mdtest0
>>>
>>> For /dev/loop[0123]
>>>
>>> Then I ran
>>>
>>>    mdadm -CR /dev/md0 -e0.90 -l0 -n4 /dev/loop0 /dev/loop1 /dev/loop2 /dev/loop3
>>>    mdadm: 0.90 metadata does not support layouts for RAID0
>>>
>>> My question is whether the regression tests in the tarball are valid for mdadm-4.2?
>>>
>>>    -- Bruce Dubbs
>>>       linuxfromscratch.org
>>>
>>> Note: The kernel is version 5.15.12.
>>
> 

Re: mdadm regression tests fail

[Mariusz Tkaczyk]

On Wed, 5 Jan 2022 14:42:31 -0600
Bruce Dubbs <bruce.dubbs@gmail.com> wrote:

> My point is that many of the tests fail.  It's not that someone
> should use the superblock v0.9.  That's only an example. The test
> should be removed or marked "Expected FAIL" or similar.  Our users
> run the tests as a confidence check that the build is successful.
> They are generally not trying to debug the package.
> 
> I can certainly say that the tests are broken and leave it at that.
> If it were only a couple of tests that fail, we generally say
> something like testA and testG are known to fail, but in this case
> fully half of the tests fail.
> 
> I would like to know what the maintainers think of the regression
> tests.  Are they maintained?  Should they all pass?  For our users
> there are far too many tests to run them individually.
> 
>    -- Bruce
> 

Hi Bruce,
I can say that at least IMSM test are maintained and used regularly.
You can use this subgroup. I can also see some test
improvements submitted last time in mdadm repository, so I can assume
that there are some usages outside IMSM (at least for 1.x metadata).

Without continuous integration testing in upstream it is externally hard
to have all tests in good shape. The verification is done by users, as
you can see it is not used frequently. I'm closer to say that this part
in not maintained. You can take a challenge and fix them :)

Thanks,
Mariusz