On Sun, Feb 03, 2002 at 01:13:40AM +0100, Matthias Andree wrote:
> > Though I use to make write cache enabled, but in this 'test' I turn
> > it off.
>
> That's important for production use also!

Yeah of course. I did the test on my own computer at home. So,
If anybody suggest me to throw some water to it instead of just
push reset button, I'll doit :-))) <*kidding*>.

Yes I disabled write cache.

root:~> hdparm -I /dev/discs/disc0/disc
...
Commands/features:
        Enabled Supported:
           * READ BUFFER cmd
           * WRITE BUFFER cmd
           * Host Protected Area feature set
           * look-ahead
                write cache
           * Power Management feature set
                Security Mode feature set
           * SMART feature set
           * DOWNLOAD MICROCODE cmd
...

> It's visible from mailing list archives, I once asked Andrew Morton
> about ext3 sync behaviour.

Yes I read about that to. Thanks for taking up this issue.

> Presumably, it would, however, any fsync() after the link() would also
> flush the link() with Linux 2.4's ext3fs

hmm.. sounds like a wish list to me :-) Would add some #ifdef on
sane_link is enough? I use hdparm -W0 and chattr -S on postfix
queue. I don't see any problem (with this test). But asynchronous
link(2) keeps bothering me. I use ext3fs.

I did chattr -S on postfix's queue, link libsyncdir against local
daemon only and did the test again several times. No message lost
(though I got more than 1000 mails on my maildir and several files
on Maildir/tmp). It's okay.

I think this maildir delivery workaround is worth of effort ...

Thanks.

Regards,

P.Y. Adi Prasaja

-
To unsubscribe, send mail to majordomopostfix.org with content
(not subject): unsubscribe postfix-users

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]