[kwlug-disc] eSATA hot pluggable on Linux?

[unsolicited]

Khalid Baheyeldin wrote, On 06/01/2010 7:07 PM:
> On Tue, Jun 1, 2010 at 11:28 AM, Raul Suarez <rarsa at yahoo.com> wrote:
> 
>> --- On *Tue, 6/1/10, Khalid Baheyeldin <kb at 2bits.com>* wrote:
>>
>> Anyone had experience with eSATA and is it hotpluggable under Linux?
>> Do you need certain BIOS features for this?
>>
>> I have an eSATA bay for my external USB/eSATA HDD drive.
>>
>> I haven't found a way to make it hot swap. If I have to remove it or
>> re-insert it, I must turn off my computer.
>>
> 
> That was exactly what I was seeing: the disk had to be plugged in with the
> computer turned off to be identified.
> 
> Unsuitable for a server of course.

Depends on the server I suppose - if you have RAID SCSI, probably a 
hardware card, you'd probably go to RAID SATA, and a hardware card, 
and no more expect hotswap with it than you would with SCSI. Both 
probably better (more successfully) implemented via hardware than 
within the OS. You'd probably also be doing replicated over the fibre 
channel / ethernet network replications, not hot swap. Particularly 
with VMs. Heck, even with VMs, this issue is back - on the fly 
replicated VMs is neither trivial nor cheap.


The same was supposed to be true of SCSI - hot pluggable. And the sata 
code was based on that, IIRC. And SCSI is still not there yet either. 
IIRC. [I suspect most successful hot swap SCSI drives have more to do 
with smarts on the raid card, than successful implementation within 
the OS.]

So why is it considered reasonable to have to shut down to change 
SCSI, or IDE, but not external SATA (which is actually exactly the 
same as internal SATA, with a bit more mechanical 'glue'.)

Agreed - it would be nice if ...

And, I know it'll come. Wouldn't have had that level of faith, pre-SATA.

Now, using the Windows world for reference, all devices were happier, 
be it SD, USB, or whatever, if you told it to stop before actually 
yanking the hardware. I don't think the requirement to tell it to 
disconnect, before yanking it, unreasonable. The problem in the 
Windows world is that traditionally drives, internal or external, SATA 
or IDE, had no mechanism for the user to ask it to be disconnected. I 
have had some, inconsistent, success with eSata in the Windows world.

I wonder, given the analogy, if you wouldn't be more successful 
(currently) using NAS over gigabit, using SATA drives. I'm thinking 
the NAS might have the hotswap smarts in it to make it more successful 
in this area, much as the raid cards over the OS, appears to. Heck, 
even a SATA NAS equivalent, connected via SATA, with onboard 
controller in the device, might be more successful. The SATA hotswap 
spec being more effectively implemented than that within the OS?