Mailinglist Archive: opensuse (2324 mails)

[Anders Norrbring <lists@xxxxxxxxxxxx>]

Anders Norrbring skrev:
> Per Jessen skrev:
> > Anders Norrbring wrote:
> >
> > > It's really not a big deal since the SAS array is RAID-5 with 2 hot
> > > spare drives, so I'm pretty certain drive failures won't be a problem,
> >
> > You meant to say "I'm certain drive failures will be a problem" ? :-)
> >
> > A key issue with a RAID1 or -5 is the time it takes to recover.  Whilst
> > the array is recovering, you're very much exposed.  A RAID5 built of
> > 1Tb drives could take quite a while to recover from a one drive
> > failure.  One way of reducing the risk is to use drives with staged
> > manufacturing dates, i.e. not all from the same batch/date.
> >
> >
> > /Per
>
>
> Haha.. ;-) Well, we all know drive failures WILL happen at some time.. 
> But with a hardware SAS RAID-5 and 2 hot spares, I don't really see that 
> it would create any acute problems besides possibly the rebuild time 
> which is around 2 hours for a failed drive in my case.
> And yes, all the 8 drives are from different batches and dates.
>
> That's not really the issue here, I was more thinking of a "hot mirror 
> backup" of the virtual machines, but as I mentioned, I'll do some 
> experimenting with the speeds. I'll post back my findings when done.
>
> Anders.


I've been running bonnie++ 5 times on each individual drive and on the 
mirrored volume, and from what I can see, the mirror set is slightly 
slower than the fastest drive (about 85% of the single drive speed) but 
not close to as slow as the slower drive.
My conclusion is that the software RAID in fact do mirror/buffer in the 
background to keep the speed up.

Anders.
--
To unsubscribe, e-mail: opensuse+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: opensuse+help@xxxxxxxxxxxx