[kwlug-disc] eSATA hot pluggable on Linux?

[unsolicited]

Khalid Baheyeldin wrote, On 06/01/2010 8:42 PM:
> On Tue, Jun 1, 2010 at 8:07 PM, unsolicited <unsolicited at swiz.ca> wrote:
> 
>> 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'.)
>>
> 
> Because this is a hard disk dock that works with USB for removable disks,
> and is supposed to work with SATA too.  It has the connector for SATA, and
> provided a bracket and a cable as well.
> 
> http://www.vantecusa.com/en/product/view_detail/295
> 
> The reason I want SATA over USB is that the former is faster, and less CPU
> intensive as well.
> 
> The whole idea is to unmount a disk, power off the dock, and change the
> disk. Works fine with USB, but not SATA (on my server).

Oh I get all this. Didn't you recently pick up on eSATA from me? (Some 
months ago.)

> The issue here is probably that the motherboard does not support the SATA
> port multiplier feature
> http://www.serialata.org/technology/port_multipliers.asp

Actually ... isn't it the motherboard's sata chipset that matters? - 
Everything else in Linux seems to depend upon chipset! (-:

Which is to say ... would a cheap sata pci adapter, with an 
appropriate chipset, and port on bracket, rather than an entire new 
computer, be expeditious?

I have several, from CanadaComputer, and been quite pleased. But not 
using under Linux at the moment.

Also got an external enclosure. I think it took a whole 2 days before 
I separated the faceplate (guts) from the casing, and now just walk 
around with the bare drive and plug it in wherever I need it. (I think 
I actually accidentally bought a 2.5" enclosure, then discovered my 
drives were 3.5" and wouldn't fit. <sigh>) I haven't looked back, and 
never noticed the difference. It ain't pretty - but I don't care. 
Looks geeky, and that may be no bad thing.

Note - by extension - the drive doesn't have to be on your server. 
Given today's drive capacities, you can probably back up multiple 
computers to one drive. So set your scripts up, replicating to a drive 
off your desktop (at night), leave your server alone, and just deal 
with the desktop when and as you need to. [This is why I have gigabit 
throughout the house, and why those who say you don't need a gigabit 
gateway are wrong.] By extension - a real 'bad' thing, bad filesystem 
driver, electrical spike, is less likely to take your backup out as a 
result of the same event. As you note on your web page - it doesn't 
matter if it takes 4 hours directly, or 8 hours over your home 
network, when it's run overnight. And you'll be less distressed, 
perhaps not by much, at having to take your desktop down to detach 
your drive, than the server.

By the same reasoning - it is arguable that you backup to another 
internal drive on another machine than the server within the house, 
then backup from there to the external drive. Then it doesn't matter 
if the over the net transfer takes up to 24 hours. And you gain 
on-premise, one day apart, redundant backups. In your current 
situation - for only the cost of one additional drive. You have 
everything else already. If that internal drive is SATA - if 
circumstances are appropriate, you pull the drive, put it in your 
external, and proceed. e.g. Restoring from failure prior to having net 
available.

Finally, it is further arguable ... having an eSata drive, JUST before 
you put a new machine into production, you locally Clonezilla off a 
disk image of it to the eSata. The biggest PITA and time loss is just 
getting a recovered machine up and going to the point of being able to 
worry about what you did to the original, since initial production 
implementation. At least with this you avoid the whole OS / App. 
reinstall process.

Data paranoia is not an affliction - it's a healthy lifestyle!

They do, actually, thank you later!