fedora 16, ext4 corruption

[Ian Malone]

On 22 January 2013 17:28, Bill Davidsen <davidsen at tmr.com> wrote:
> Ian Malone wrote:
>>
>> Hi,
>>
>> Is anyone else seeing problems with F16 ext4? I recently moved my
>> install onto a new harddisc (which is reporting SMART is fine after
>> running a full test) and did that by making a new ext4 fs (previously
>> ext3) then rsyncing across. I'm now seeing occasional system 'freezes'
>> (not hard freezes, but programs can fail to start or freeze. Yesterday

> I have multiple TB running on fc16+ext4, and have not seen any problems of
> any kind with the filesystem. I do occasionally see a hang, this happens
> with rsync if you copy something large to a machine with significant memory,
> at the end rsync does a sync() I believe (if it was doing fsync() I think
> the pain would be spread) and the system is emptying the i/o buffers to
> disk. A typical single disk will write at 50-100MB/s and if you have 4GB or
> more of backed up data, it will get slow. There are disk tuning tips you can
> use to spread the pain, but it makes the copy of small files run slower in
> many cases.
>
> If you have a lot of small files, your journal may be full of incomplete
> transactions. The solution is to make the journal larger, I don't know if
> modifying the journal after the fact will change that, or even if it's the
> problem. I have played with putting journal on a faster device, 10k rpm
> drive, SSD, or memory (yes, just for testing, I know it's unsafe). Faster
> journal makes things faster with many small files.
>
> Lots of thoughts for you to check against your particular problems.
>

Any way to check the journal for incomplete transactions? Except for
immediately after a crash the parition comes up clean in an e2fsck
from a livecd. It made three clean passes through memtest, so I don't
think that's the issue.


-- 
imalone
http://ibmalone.blogspot.co.uk