XFS write cache flush policy
Matthias Schniedermeyer
ms at citd.de
Thu Dec 20 04:18:23 CST 2012
On 19.12.2012 23:48, Stan Hoeppner wrote:
> On 12/19/2012 7:40 PM, Matthias Schniedermeyer wrote:
>
> > I had thought about issuing a
> > 'sync', because the previous incarnation of the bug flashed before my
> > inner eye. But by the time i had shutdown X and got to the
> > command-prompt i had already forgotten to type 'sync' and went straight
> > for 'reboot'
>
> When did you last read the man pages for reboot and shutdown? reboot
> issues a sync automatically unless you specify
>
> -n Don't sync before reboot or halt. Note that the kernel and
> storage drivers may still sync.
>
> (note the disclaimer)
>
> and even though the shutdown man page doesn't state so explicitly, I'm
> pretty darn sure that shutdown issues a sync before halting or rebooting.
>
> Thus, you shouldn't have to manually call sync before issuing your
> favorite system restart command. If sync isn't being called, maybe
> there's an issue with the versions of these commands shipped with your
> distro, or other type of problem.
That's strange, i looked into the source of /sbin/halt and it does
contain a call to 'sync'.
So it either didn't work, or i hit the umount-bug that David mentioned.
When i rebooted to 3.7(.0), from a buggy 3.6.2, i 'sync'ed manually
before rebooting and also manually 'umount'ed all "physical" FSes except
the rootfs and this time i didn't get a corruption. All
mount-syslog-lines for the 3.7 boot say the filesystem was clean.
So when it is time to reboot to 3.7.2, which hopefully contains the
bugfix, i still intent to do a wetware (half-)shutdown before the actual
'reboot'. The statistic is 2 out of 3 in favor of corruptions in the
last 3 reboots of this particular computer.
--
Matthias
More information about the xfs
mailing list