Sleeping harddrives?

Sleeping harddrives?

[Anders Widman]

   How  would  ReiserFS  react  when  a user/program/anything tries to
   access a file on a harddrive that is spin down?

   I  remember that I tested this about a year ago, and it didn't work
   very well. The filesystem timeout was shorter than the time to spin
   up the drive.

   //Anders

Re: Sleeping harddrives?

[Oleg Drokin]

Hello!

On Sun, Jun 02, 2002 at 02:34:06AM +0200, Anders Widman wrote:

>    How  would  ReiserFS  react  when  a user/program/anything tries to
>    access a file on a harddrive that is spin down?

Linux IDE/SCSI layer issupposed to spind the drive up.

>    I  remember that I tested this about a year ago, and it didn't work
>    very well. The filesystem timeout was shorter than the time to spin
>    up the drive.

Huh? What filesystem timeout?

Bye,
    Oleg

Re: Sleeping harddrives?

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 1520 bytes --]

On Sun, 02 Jun 2002 21:49:27 +0400, Oleg Drokin <green@namesys.com>  said:
> On Sun, Jun 02, 2002 at 02:34:06AM +0200, Anders Widman wrote:
> >    I  remember that I tested this about a year ago, and it didn't work
> >    very well. The filesystem timeout was shorter than the time to spin
> >    up the drive.
> Huh? What filesystem timeout?

I think Anders hit a timing hole.

Quite often, high-end drives have a "delay spin-up based in SCSI ID" feature,
so all the drives numbered 0 spin up immediately, all the drives numbered 1
spin up after 5 seconds delay, all the drives numbered 2 wait 10 seconds, etc.
This is to prevent the spin-up surge from blowing out the power supply if
a very large RAID array all tries to spin up at once (the current needed to
accelerate the drive to 10K RPM is on the order of 3-4 times what is required
to KEEP it spinning at that speed).

Unfortunately, if you're drive number 14, you have a 70 second wait before you
spin up - and if your boot devices was on drive 0 (and quite possibly on a
previously-spun-up drive), it's quite possible that the system will boot and
manage to fsck the lower-numbered drives before you spin up.  So you look like
a "dead" drive... Whoops.

I suspect the same thing could happen with a hot-swap SCSI drive - plug it
in, run to the console, and type "fsck" or "mount" before the drive has spun up
(especially if as before, it's drive 14 and set to delay spin up).

-- 
				Valdis Kletnieks
				Computer Systems Senior Engineer
				Virginia Tech




[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: Sleeping harddrives?

[Anders Widman]

> Hello!

> On Sun, Jun 02, 2002 at 02:34:06AM +0200, Anders Widman wrote:

>>    How  would  ReiserFS  react  when  a user/program/anything tries to
>>    access a file on a harddrive that is spin down?

> Linux IDE/SCSI layer issupposed to spind the drive up.

It  is  supposed  to,  yes.  When I tried this before, it did spin the
drives up, but didn't wait enough before trying to access it, so I got
strange errors.

My question is simple. Is this something that still occur today?

>>    I  remember that I tested this about a year ago, and it didn't work
>>    very well. The filesystem timeout was shorter than the time to spin
>>    up the drive.

> Huh? What filesystem timeout?

Well, it might not even be the filesystem.


//Anders

> Bye,
>     Oleg

Re: Sleeping harddrives?

[Oleg Drokin]

Hello!

On Sun, Jun 02, 2002 at 10:56:51PM +0200, Anders Widman wrote:
> >>    How  would  ReiserFS  react  when  a user/program/anything tries to
> >>    access a file on a harddrive that is spin down?
> > Linux IDE/SCSI layer issupposed to spind the drive up.
> It  is  supposed  to,  yes.  When I tried this before, it did spin the
> drives up, but didn't wait enough before trying to access it, so I got
> strange errors.
> My question is simple. Is this something that still occur today?

At least it works for me on IDE.

Bye,
    Oleg