Topic: eSATA?

[dmay]

Does anyone have an experience with eSATA controllers and external hdd enclosures (either single or multiple drives) on SME 7?

Personally I'm interested in seeing how effective an eSATA controller and external removable hdd enclosure could be used as an effective backup device. Swapping removable hard drives on a daily basis. I know there are a lot of people doing this with USB2 setups. Just wondering if anyone has tried an eSATA setup.

Darrell

[wjhobbs]

Hi Darrell,

I wanted to try the same thing. However, although the hardware should be OK with hotswap, my understanding is that RedHat/Centos does not have any support for SATA hot-swap.

I'd prefer not to have to take the server down just to change external backup drives.

I must admit that I haven't tried it to see what bad things happen.

John

[dmay]

...have any support for SATA hot-swap.

Hi John.

In this case hot-swap is not required. The drives are not active system drives nor intended to be part of a RAID scenario. These are to be secondary where we simply need to mount, perform a backup, dismount, remove and replace. Then next backup cycle everything is repeated.

Darrell

[wjhobbs]

Hi Darrell,

I could be wrong (have been so often before) but my understanding is that there is some risk in the "dismount, remove and replace" part of things, even if the drive has been unmounted.

With USB, the system detects when a new drive has been connected and does something like a probe to see what's what. It is my understanding that this does not happen (yet) with SATA and there are issues if the new drive is not identical (physical geometry etc.) with the drive just removed, because the values from the original probe are not refreshed. (This is my non-technical understanding.)

If you find that is not the case, I would really like to know because I want to use that facility. (I've already purchased an external USB/eSATA enclosure.)

John

[dmay]

and there are issues if the new drive is not identical

Think tape drive and tape media. You would always use the same media. Here the media is a hdd. Same logic applies, use identical hdds. Since you state you have an eSATA enclosure, give this a test and report your results.

Darrell

[piran]

External SATA (eSATA):
http://www.sata-io.org/esata.asp

Not to be confused with Esata (creole English):
http://www.beginnersgame.com/esata.html
or Expatriate & Self  Assessment  Tax  Advisers -  ESATA.CO.UK

A viewpoint from the United Kingdom's shores...
Existing server does not possess an eSATA compliancy
but the motherboard does have a spare PCI-X slot.
Difficulty is going to be deriving a suitable adapter card
and its compliancy with SME7. The cabling and enclosures
look to be an easier proposition, with hot-swap a possibility.
Possible source of eSATA adapter cards (bottom of page 'Tempo'):
http://www.sonnettech.com/product/pci_adptrs.html
http://www.sonnettech.com/product/tempo-x_esata8.html
http://sonnettech.com/news/press/pr2005/pr041405_hotswap.html
Possible UK supplier of eSATA PCI-X adapter card:
http://equip-u.com/shop/index.php?main_page=product_info&manufacturers_id=9&products_id=873
...notice the complete absence of any nods whatsoever to Linux ;~/

Possible SME7/Linux compliancy with adapter card & enclosure kit...
http://www.afterhours.co.uk/highpoint-esata-upgrade-kit-v2-p-55.html
...at UKP35 for the kit it seems very cheap and cheerful.
Any experience with this hardware source?
Is it going to be likely that it is SME7 compliant?
Forget it, I've just quickly read through...
http://www.dansdata.com/esata.htm
...and I've gone off the boil with the Rocket stuff.

MacUpgrades eSATA FAQ...
http://www.macupgrades.co.uk/store/faq-esata.php
...it's very MAC-orientated but implies hotswapping works well/easily.

[dmay]

Simplest way would be to try to use a spare internal SATA connection from your existing motherboard and make it external:

http://www.satagear.com/221008A_SATA_Cable.html

Then connect an external SATA removable enclosure:

http://www.satagear.com/SS-302i-Silver_SATA_II_Product.html

Darrell

[piran]

Had gone that way of thinking already;~)
Here's a UK supplier for a short internal/external hookup:
http://www.starmount.co.uk/productversion/413.html

Been thinking... I have an even simpler way that might be feasible.
SME7 is running a h/w RAID card (PCI-X) and only 5 out 8 connectors
are in use due to an unpublished (2TB) limitation restricting the number
of 400GB drives I fitted. There's a 6th one in there as a cold standby.
The h/w RAID BIOS is capable of setting up another RAID or even a
JBOD with the remaining 3 connectors. Don't know for sure whether
the 2TB limitation was for a single iteration of RAID (!?) or whether
it's a limitation of the h/w RAID card and/or its BIOS. The connx are
5cm away from the backplane's L-plates and a full set of eight cables
(admittedly 'internal' spec) were supplied. What about just a simple
inside outside hookup? Physically it's all safe enough (closed zone),
the server can't move as its PSU fan outlet is hotglued to thermally
lagged ventilation ducting. EMI? Will need a longer DC harness,
though I'd prefer an external power pack as the resident PSU is
quite probably max'd out. At least I wouldn't have to worry about
a compliant PCI-X eSATA card;~)

[wjhobbs]

From dmay:

Since you state you have an eSATA enclosure, give this a test and report your results.

This is the testing I did and conclusions I reached.

Hardware
I acquired the following equipment and used it for this test:

Vantec UGT-ST300 SATA/eSATA PCI Host Controller (150MB/s)
Kingwin KF-813-BK Mobile Rack for SATA HDD
Kingwin KF-813-T-BK Inner Tray for KF-813-BK (The 813 model has two fans in the tray and a fan in the rack to promote cooling.)
Maxtor 300GB SATA-150 (2 drives)
SATA data cable

This PCI card has an internal and an external (eSATA) connector. This configuration uses the internal port to connect to a removable drive bay. A connection through the external port to an eSATA enclosure should work identically; but I did not test this.

Setup
Installed a drive in the tray that came with the rack and the other disk in the extra tray.
Installed the controller in a PCI slot and installed the rack in a 5.25” bay, attached a standard power connector, and used the SATA cable to connect the controller to the rack.

Testing
Booted the machine.
Inserted one tray into the rack and secured it with the key (drive power light on).
The drive was not visible to the system. Originally thought the driver for the SATA controller had not been loaded; but the message log indicates that the card was probed but no device was found on either channel.

Code: [Select]

SCSI subsystem initialized
libata version 1.20 loaded.
sata_sil 0000:02:06.0: version 0.9
ACPI: PCI interrupt 0000:02:06.0[A] -> GSI 11 (level, low) -> IRQ 11
ata1: SATA max UDMA/100 cmd 0xF8840080 ctl 0xF884008A bmdma 0xF8840000 irq 11
ata2: SATA max UDMA/100 cmd 0xF88400C0 ctl 0xF88400CA bmdma 0xF8840008 irq 11
ata1: no device found (phy stat 00000000)
scsi0 : sata_sil
ata2: no device found (phy stat 00000000)
scsi1 : sata_sil
Rebooted the machine with drive connected.
Drive visible.
Created partition and formatted as ext3.
Performed dar2 backup (which includes mounting and unmounting partition).
Removed drive.
Inserted second drive.
Drive visible.
Created partition and formatted as ext3 with identical parameters as for the first drive.
Performed dar2 backup.
Removed drive.

Initiated a reboot.
Error messages on shutdown indicated that scsi was having problems syncing cache. Clearly the system thought the drive was still there. The system finally concluded the reboot.
Drive not visible.
Inserted tray into the rack and secured it with the key (drive power light on).
Drive not visible to system.
Removed this drive and inserted first drive in system.

Reboot with drive power light on.
Drive visible, mountable and contains the original backup dar2 data files.

Conclusions
The lack of basic “hot swap” capability in the current kernel means adding and removing a drive will not be detected.
Therefore:

- the drive must be attached and active at boot
- any subsequent replacement must be with an identical drive
- there is a risk of significant problems if the drive is disconnected before it has been unmounted
- the drive should be attached and active at shutdown or reboot

If these constraints are respected, it MAY be possible to take advantage of the removable characteristics of a SATA drive.
Given the backup performance I described in this posting http://forums.contribs.org/index.php?topic=34041.msg148037#msg148037 there are advantages to doing so.

HOWEVER, this is not supported! You do so at your own risk!
 
John

[piran]

An admirable report John;~) Did you notice or otherwise record whether
the backup to eSATA resulted in rampant CPU (1 or 2) hogging or not?

[wjhobbs]

Piran,

I did not look specifically at CPU usage (it's a single cpu box). However, while the backup was going on I accessed server manager and several files in different ibays and did not see a noticable degradation in response time.

John

[dmay]

Great report John!

there is a risk of significant problems if the drive is disconnected before it has been unmounted

Correct. It should also be noted that the dar2-backup action performs a lazy unmount.

Code: [Select]

-l     Lazy  unmount.  Detach  the  filesystem  from the filesystem hierarchy now, and cleanup all references to the filesystem as soon as it is not busy anymore. This  option  allows  a  'busy'  filesystem  to  be  unmounted.
This permits the script to exist cleanly on a busy system and the drive to be unmounted as soon as the system has a chance. In a typical backup environment this works well as backups are usually scheduled for overnight processing. The drive and system will have lots of time to complete the unmount. However if you are testing as was your case and immediately removing the drive at job completion it may not have had time to fully complete the unmount process. This may be where your scsi syncing issues arose from during reboot.

Darrell

[zeeclor]

I am not sure I am in the right thread but I am having a problem with my Stardom e-sata (http://www.raidon.com.tw/web/pro-sr3610.htm) RAID 1 and sme 7.0. It works fine under debian and edgy but anaconda dies quickly despite loading the via_sata driver for my MSI board using a VIA 8237A south bridge chipset.

I know (now) it's not supported upstream but it is a shame since it would have been a nice solution to hot backup and failover in my small to medium enterprise.

David