Re: Bad ext4 sync performance on 16 TB GPT partition

[Eric Sandeen <sandeen@redhat.com>]

Karsten Weiss wrote:
> Hi Eric!
> 
> On Mon, 1 Mar 2010, Eric Sandeen wrote:
> 
>>> Sorry for the delay, here's the (good) 2.6.32 result:
>>>
>>> # /usr/bin/time bash -c "dd if=/dev/zero of=/mnt/large/10GB bs=1M count=10000 && sync"
>>> 10000+0 records in
>>> 10000+0 records out
>>> 10485760000 bytes (10 GB) copied, 46.3369 seconds, 226 MB/s
>>> 0.00user 14.17system 0:59.53elapsed 23%CPU (0avgtext+0avgdata 6224maxresident)k
>>> 0inputs+0outputs (0major+1045minor)pagefaults 0swaps
>>>
>>> To summarize:
>>>
>>> Bad:  2.6.18-164.el5 (CentOS)
>>> Bad:  2.6.18-164.11.1el5 (CentOS)
>>> Bad:  2.6.18-190.el5 (RH)
>>> Good: 2.6.32
>>> Good: 2.6.33
> 
> In the meantime I've also reproduced the problem on another machine with a 
> Red Hat 5.5 Beta (x86_64) installation and decided to open a bug on RH's 
> bugzilla:
> 
> Bad ext4 sync performance on 16 TB GPT partition
> https://bugzilla.redhat.com/show_bug.cgi?id=572930

Thanks, and thanks for double-checking upstream.

>> Thanks, I'll have to investigate that.  I guess something may have gotten lost
>> in translation in the 2.6.32->2.6.18 backport.....
> 
> Did you come up with anything I could test?

I'll look into this...
 
> Is anyone else able to reproduce the problem?

Since it's not an upstream problem, this issue is probably best discussed
in the RHEL bug, now.

Thanks,
-Eric