Slowness on ext4 master branch (Was: The new ext4 git tree re-org)

Slowness on ext4 master branch (Was: The new ext4 git tree re-org)

[Amir Goldstein]

On Wed, May 11, 2011 at 9:59 PM, Amir Goldstein <amir73il@gmail.com> wrote:
> On Tue, May 10, 2011 at 4:56 PM, Theodore Ts'o <tytso@mit.edu> wrote:
>>
>> I've uploaded an updated dev branch to the ext4 tree.  Compared to the
>> previous version of dev branch, I've removed my modified version of
>> Allison's "NULL pointer when make_indexed_dir returns -ENOSPC" patch,
>> and I've added Amir's alloc_semp removal patches.  I'm still in the
>> process of reviewing those patches; I've included them to make it easier
>> for I and others to test them.
>>
>
> I have a general question regarding testing for next merge window.
> What kernel do you usually use when testing the dev branch?
> I started writing an xfstest to test online resize and filesystem
> filling in parallel, based on test 224 (Delayed allocation at ENOSPC test).
> I started running test 224 on a kernel built from ext4/master
> and it seemed to run too slow. looking at the waiting tasks
> it looked like it could be related to I/O less throttling.
> Since ext4/master is based on 2.6.39-rc3, I figured maybe I am running
> a buggy kernel, so I merged 2.6.39-rc7 to ext4/master to build a new kernel.
>
> Am I going in the right direction at all?
> I have a dedicated server for tests (X64_64 quad core), so it you would
> like me to run some xfstest runs with specific configuration, or certain
> patches applied, please let me know.
>
>

So now that I know I am not hearing voices, I guess test 224 is also too new
to be tested by many people.
The good news is that test 224 runs smoothly on kernel 2.6.39-rc7.
on ext4 master branch is either hung or took too long time for me to wait.
I will try to find the offending commit on Monday.

Amir.
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Slowness on ext4 master branch (Was: The new ext4 git tree re-org)

[Amir Goldstein]

On Sat, May 14, 2011 at 9:49 PM, Amir Goldstein <amir73il@gmail.com> wrote:
> On Wed, May 11, 2011 at 9:59 PM, Amir Goldstein <amir73il@gmail.com> wrote:
>> On Tue, May 10, 2011 at 4:56 PM, Theodore Ts'o <tytso@mit.edu> wrote:
>>>
>>> I've uploaded an updated dev branch to the ext4 tree.  Compared to the
>>> previous version of dev branch, I've removed my modified version of
>>> Allison's "NULL pointer when make_indexed_dir returns -ENOSPC" patch,
>>> and I've added Amir's alloc_semp removal patches.  I'm still in the
>>> process of reviewing those patches; I've included them to make it easier
>>> for I and others to test them.
>>>
>>
>> I have a general question regarding testing for next merge window.
>> What kernel do you usually use when testing the dev branch?
>> I started writing an xfstest to test online resize and filesystem
>> filling in parallel, based on test 224 (Delayed allocation at ENOSPC test).
>> I started running test 224 on a kernel built from ext4/master
>> and it seemed to run too slow. looking at the waiting tasks
>> it looked like it could be related to I/O less throttling.
>> Since ext4/master is based on 2.6.39-rc3, I figured maybe I am running
>> a buggy kernel, so I merged 2.6.39-rc7 to ext4/master to build a new kernel.
>>
>> Am I going in the right direction at all?
>> I have a dedicated server for tests (X64_64 quad core), so it you would
>> like me to run some xfstest runs with specific configuration, or certain
>> patches applied, please let me know.
>>
>>
>
> So now that I know I am not hearing voices, I guess test 224 is also too new
> to be tested by many people.
> The good news is that test 224 runs smoothly on kernel 2.6.39-rc7.
> on ext4 master branch is either hung or took too long time for me to wait.
> I will try to find the offending commit on Monday.

I couldn't reproduce this.

>
> Amir.
>
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Slowness on ext4 master branch (Was: The new ext4 git tree re-org)

[Eric Sandeen]

On 5/14/11 1:49 PM, Amir Goldstein wrote:
> On Wed, May 11, 2011 at 9:59 PM, Amir Goldstein <amir73il@gmail.com> wrote:
>> On Tue, May 10, 2011 at 4:56 PM, Theodore Ts'o <tytso@mit.edu> wrote:
>>>
>>> I've uploaded an updated dev branch to the ext4 tree.  Compared to the
>>> previous version of dev branch, I've removed my modified version of
>>> Allison's "NULL pointer when make_indexed_dir returns -ENOSPC" patch,
>>> and I've added Amir's alloc_semp removal patches.  I'm still in the
>>> process of reviewing those patches; I've included them to make it easier
>>> for I and others to test them.
>>>
>>
>> I have a general question regarding testing for next merge window.
>> What kernel do you usually use when testing the dev branch?
>> I started writing an xfstest to test online resize and filesystem
>> filling in parallel, based on test 224 (Delayed allocation at ENOSPC test).
>> I started running test 224 on a kernel built from ext4/master
>> and it seemed to run too slow. looking at the waiting tasks
>> it looked like it could be related to I/O less throttling.
>> Since ext4/master is based on 2.6.39-rc3, I figured maybe I am running
>> a buggy kernel, so I merged 2.6.39-rc7 to ext4/master to build a new kernel.
>>
>> Am I going in the right direction at all?
>> I have a dedicated server for tests (X64_64 quad core), so it you would
>> like me to run some xfstest runs with specific configuration, or certain
>> patches applied, please let me know.
>>
>>
> 
> So now that I know I am not hearing voices, I guess test 224 is also too new
> to be tested by many people.

I hope not, it was committed over a year ago :)

If it does enospc, ext4 gets into the flushing & syncing mode as things
fill up, so maybe something changed in that area?

-Eric

> The good news is that test 224 runs smoothly on kernel 2.6.39-rc7.
> on ext4 master branch is either hung or took too long time for me to wait.
> I will try to find the offending commit on Monday.
> 
> Amir.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html