dd's 'direct' flag will make the kernel do the IO immediately.

Ray Higgs
System z FCP Development
Bld. 706, B24
2455 South Road
Poughkeepsie, NY 12601
(845) 435-8666,  T/L 295-8666
[email protected]

Linux on 390 Port <[email protected]> wrote on 11/23/2010 06:18:27
AM:

> Shane <[email protected]>
> Sent by: Linux on 390 Port <[email protected]>
>
> 11/23/2010 06:18 AM
>
> Please respond to
> Linux on 390 Port <[email protected]>
>
> To
>
> [email protected]
>
> cc
>
> Subject
>
> Re: Problems with DASD driver on SLES 11 SP1?
>
> "dd" is _the_ absolute worst option for backups. It is stupid, has no
> concept of data integrity, and is stupid.
> And the default conv options are appalling - and it's stunningly stupid.
>
> Great for image  copies in failure/forensic situations, other than that
> it's downright ... er, see above.
>
> [f]sync doesn't really help a lot in this situation either - from a
> Linux perspective. Last I looked (on a non-z system), fsync doesn't
> guarantee the (physical) I/O actually gets written then-and-there.
> Linux has I/O schedulers in play, and then there is pdflush/BDI to
> worry about waking up. Sometime, somewhere the I/O does get flushed.
> Blktrace has been around for a while now - the (timimg) output makes
> for interesting reading.
> Then we have z/VM snoozing away in the corner ... Sebastian and his
> cohorts can fill you in there.
>
> If it (ever) worked you were IMHO lucky. I'd be slipping some (long)
> sleeps in your scripts.
>
> Shane ...
>
> On Tue, 23 Nov 2010 10:51:53 +0100
> Sebastian Ott wrote:
>
> > > 2. Problem:
> > >
> > > Because I lack the FLASHCOPY function I use for disk to disk copy
> > > dd.
> > >
> > > dd bs=4096 if=/dev/dasdx of=/dev/dasdy
> > >
> > > In principle it works ok. But when I do a chccwdev -d 0.0.xxxx even
> > > with an udevadm settle --timeout=20 afterwards,
> > > the file system gets corrupted, when I attach it to another guest.
> > >
> > > I am really lost since all that was working well under SLES 10 SP2.
> > > I would really like to know if someone had similar experience in
> > > SLES 11 SP1.
> >
> > The data is not guaranteed to be written in this case. This is due to
> > changes in the linux block layer (it was not guaranteed to be written
> > in sles 10 either but worked by chance).
> > You have 2 options here:
> > * tell dd to sync its data (check the man page - i think the option is
> > fsync)
> > * manually run "sync" after dd returns
>
> ----------------------------------------------------------------------
> For LINUX-390 subscribe / signoff / archive access instructions,
> send email to [email protected] with the message: INFO LINUX-390 or
visit
> http://www.marist.edu/htbin/wlvindex?LINUX-390
> ----------------------------------------------------------------------
> For more information on Linux on System z, visit
> http://wiki.linuxvm.org/

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/

Reply via email to