[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#474747: mdcfg: Avoid sync fresh created arrays



Hi Ferenc Wagner!
 On Wed, May 07, 2008 at 04:02:10PM +0200, Ferenc Wagner wrote next:

> > On Mon, Apr 07, 2008 at 07:46:17PM +0200, Ferenc Wagner wrote next:
> >
> >> Andrey Melnikov <temnota@kmv.ru> writes:
> >> 
> >>> Please, add mdadm '--assume-clean' options when creating new arrays
> >>> in installer to avoid initial resync. This make huge speedups when
> >>> creating big arrays (100 Gb or more).
> >> 
> >> While I sympathize, I feel like it isn't the best solution.  You will
> >> get a huge mismatch_cnt 
> >
> > Yet antoher "mars weather device" ? 
> > Current algorithm compare all data on disks, empty sectors or dirty sectors,
> > and this counter show only internal 'mismatches'.
> 
> I don't quite follow you, but if you mean that mismatch_cnt isn't
> worth much, I accept that.
Yes. 
 
> >> and lose an initial check of your disks.
> >
> > 'SECURITY ERASE' by disk controller - very fast, and remap bad sectors if
> > found it instead of 'raid fault' detection.
> I don't know what you mean, again.
I don't need 'initial ckeck' of disks. I'm use other method to check disks
before install.
  
> >> Isn't is possible/enough to tune /proc/sys/dev/raid/speed_limit_max
> >> down for the time of the install?
> >
> > I'm talk about 'time to wait when raid complete sync zero block's from
> > one hdd to antoher'. Tuning speed_limit_max slightly decrase this time.
> Why would you have to wait?

Because, if I don't wait when resync finished and reboot server - resync
started after reboot again. Why I should wait, when kernel mirror 250Gb
zerro's from sda to sdb (also contian same zerro's)?
 
> > It's possible to add any option to use in 'expert mode' setup ?

-- 
 Best regards, TEMHOTA-RIPN aka MJA13-RIPE
 System Administrator. mailto:temnota@kmv.ru




Reply to: