
Our nightly backups outgrew the 500GB disks I was using so I've bought some 1TB disks, and they aren't working properly. Both the old and new disk are USB3 disks on a USB3 controller which seems to require a reboot every 50 plug/unplug cycles or so but otherwise previously worked well. The symptoms are: [ 7455.667194] sd 16:0:0:0: [sdj] Attached SCSI disk [ 7690.448638] EXT3-fs error (device sdd1): ext3_find_entry: reading directory #2 offset 0 [ 7690.448972] Buffer I/O error on device sdd1, logical block 0 [ 7690.448994] lost page write due to I/O error on sdd1 [ 7690.448999] EXT3-fs (sdd1): I/O error while writing superblock [ 7690.449056] EXT3-fs (sdd1): error: ext3_journal_start_sb: Detected aborted journal (the first message is an indication of when I plugged the disk in, the second is the start of it having errors) The disk itself seems okay and doesn't fail at a specific sector each time or anything, and I've tried another disk too. The USB3 cable is quite long, so I've replaced it with a shorter one and so far it's backed up 10GB since I started writing this email. On other tests it's written out around 200GB before failing. The possibilities I can think of are: . cable bad (it's well less than 3M) . usb disk incompatible with controller (USB3 controllers of that vintage are a bit hit and miss) . usb disk drawing too much power (shouldn't USB log an error about this? Are there tools to check?) . bad batch of drives (all pass smart tests) . something else Any thoughts? It's one of those problems that takes hours to test each troubleshooting step and meanwhile offsite backups are not getting done (20GB now and counting!) Thanks James

On 12/03/13 12:06, James Harper wrote:
Our nightly backups outgrew the 500GB disks I was using so I've bought some 1TB disks, and they aren't working properly. Both the old and new disk are USB3 disks on a USB3 controller which seems to require a reboot every 50 plug/unplug cycles or so but otherwise previously worked well.
The symptoms are: ... . cable bad (it's well less than 3M) ...
My household has ongoing issues with USB cables, particularly the micro ones, and it is usually power rather than signal - the specs for wire size seem to be honoured in the breech, and USB3 micro seems to elevate the problem with its dinky little contacts and less than positive mechanical locking. MSY have 1m cables with decent wire for about $6 - get a couple of those and see how you go.

On 12/03/13 13:16, Allan Duncan wrote:
On 12/03/13 12:06, James Harper wrote:
Our nightly backups outgrew the 500GB disks I was using so I've bought some 1TB disks, and they aren't working properly. Both the old and new disk are USB3 disks on a USB3 controller which seems to require a reboot every 50 plug/unplug cycles or so but otherwise previously worked well.
The symptoms are: ... . cable bad (it's well less than 3M) ...
My household has ongoing issues with USB cables, particularly the micro ones, and it is usually power rather than signal - the specs for wire size seem to be honoured in the breech, and USB3 micro seems to elevate the problem with its dinky little contacts and less than positive mechanical locking.
Seconded. I have had a nightmare month or two of failing USB cables, and new cables which can't actually push full power through themselves. I suspect this is partly because I went around upgrading my USB power sources from 500-1000 mA to 1000-2000 mA units, which is well above the original USB spec.. but yet now quite common for charging phones and driving embedded boards. I'm currently having some success with 30cm micro-usb cables from Korea that seem to work fine. I also have some $10 cables from Jaycar which are annoyingly long (1.8m) but which otherwise seem to be built well enough to carry power.

On Tue, 12 Mar 2013, Toby Corkindale <toby.corkindale@strategicdata.com.au> wrote:
Seconded. I have had a nightmare month or two of failing USB cables, and new cables which can't actually push full power through themselves. I suspect this is partly because I went around upgrading my USB power sources from 500-1000 mA to 1000-2000 mA units, which is well above the original USB spec.. but yet now quite common for charging phones and driving embedded boards.
I'm currently having some success with 30cm micro-usb cables from Korea that seem to work fine. I also have some $10 cables from Jaycar which are annoyingly long (1.8m) but which otherwise seem to be built well enough to carry power.
http://etbe.coker.com.au/2013/01/24/power-supplies-wires/ I've done some calculations about the resistance of wires for USB chargers. I had to make some assumptions about the thickness of the conductors in the cables, but anyone who's interested can make their own calculations with different assumptions. -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

On Tue, 12 Mar 2013, Toby Corkindale <toby.corkindale@strategicdata.com.au> wrote:
Seconded. I have had a nightmare month or two of failing USB cables, and new cables which can't actually push full power through themselves. I suspect this is partly because I went around upgrading my USB power sources from 500-1000 mA to 1000-2000 mA units, which is well above the original USB spec.. but yet now quite common for charging phones and driving embedded boards.
I'm currently having some success with 30cm micro-usb cables from Korea that seem to work fine. I also have some $10 cables from Jaycar which are annoyingly long (1.8m) but which otherwise seem to be built well enough to carry power.
http://etbe.coker.com.au/2013/01/24/power-supplies-wires/
I've done some calculations about the resistance of wires for USB chargers. I had to make some assumptions about the thickness of the conductors in the cables, but anyone who's interested can make their own calculations with different assumptions.
Is there anything you haven't blogged about? The USB3 cable I was using originally (the ~1.5M length one) was about the thickness of a 10A 240V power cable, although that's the outside diameter and I don't want to cut it open to find out the diameter of the wires inside. The one I'm using now is about 20cm long and is much thinner. I'm more interested in getting readings from the USB controller about the power budget and the draws of the attached devices (even though these are programmed values not actual values) in an easy to read form. James

On Tue, 12 Mar 2013, James Harper <james.harper@bendigoit.com.au> wrote:
Is there anything you haven't blogged about?
I have about 40 posts in my drafts folder... ;)
The USB3 cable I was using originally (the ~1.5M length one) was about the thickness of a 10A 240V power cable, although that's the outside diameter and I don't want to cut it open to find out the diameter of the wires inside. The one I'm using now is about 20cm long and is much thinner.
Wow, that's really thick. The conductors probably wouldn't be any thinner as there is much less need for insulation on low voltage wires.
I'm more interested in getting readings from the USB controller about the power budget and the draws of the attached devices (even though these are programmed values not actual values) in an easy to read form.
I've seen an Apple PowerBook give some information on that. I'm not sure whether it was rounded to the nearest 500mA or whether it was actually measured. It was at a LUV Beginner SIG meeting, maybe the PowerBook owner in question can comment here. -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

On 12/03/13 12:06, James Harper wrote:
. something else
Kernel bug (not saying it is here, just that it's always a possibility). cheers, Chris -- Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC

On 12/03/13 12:06, James Harper wrote:
. something else
Kernel bug (not saying it is here, just that it's always a possibility).
The reason I haven't made that assumption is that the previous 500GB disks still work fine. Any idea how to easily report on USB power usage? It wouldn't be the first time a device has lied about its current draw or an adapter has lied about its ability to supply an amount of power. James

James Harper <james.harper@bendigoit.com.au> wrote:
On 12/03/13 12:06, James Harper wrote:
. something else
Kernel bug (not saying it is here, just that it's always a possibility).
The reason I haven't made that assumption is that the previous 500GB disks still work fine.
If they're USB 3 drives as well, then that would tend to suggest it isn't a kernel bug. If they're USB 2, then one could suspect a USB 3-specific driver issue.
Any idea how to easily report on USB power usage?
No, but try mount -t debugfs debugfs /sys/kernel/deubg and then look at /sys/kernel/debug/usb/devices

On 12/03/13 15:04, Jason White wrote:
If they're USB 3 drives as well, then that would tend to suggest it isn't a kernel bug. If they're USB 2, then one could suspect a USB 3-specific driver issue.
It would also be interesting to test both a USB2 drive in a USB3 port and a USB3 drive in a USB2 port as they should be cross compatible.. cheers, Chris -- Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC

On 12/03/13 13:23, James Harper wrote:
On 12/03/13 12:06, James Harper wrote:
. something else
Kernel bug (not saying it is here, just that it's always a possibility).
The reason I haven't made that assumption is that the previous 500GB disks still work fine.
Any idea how to easily report on USB power usage?
It wouldn't be the first time a device has lied about its current draw or an adapter has lied about its ability to supply an amount of power.
You aren't kidding! I've got an Hitachi 1TB USB3 plugged in to a USB2 port, and lsusb reports: idVendor 0x4971 SimpleTech idProduct 0x1014 bcdDevice 0.00 iManufacturer 10 HitachiGST iProduct 11 Touro Mobile 3.0 ... MaxPower 8mA ... Device Status: 0x0001 Self Powered

On Tue, 12 Mar 2013, Allan Duncan wrote:
It wouldn't be the first time a device has lied about its current draw or an adapter has lied about its ability to supply an amount of power.
You aren't kidding! I've got an Hitachi 1TB USB3 plugged in to a USB2 port, and lsusb reports: idVendor 0x4971 SimpleTech idProduct 0x1014 bcdDevice 0.00 iManufacturer 10 HitachiGST iProduct 11 Touro Mobile 3.0 ... MaxPower 8mA ...
Ooh, impressive! Must use superconducting magnets for the motors and maglev bearings. How do you go about supplying the liquid helium though? Obviously your cryopump isn't USB powered. Did you have to hook up 3 phase power? -- Tim Connors

On 13/03/13 11:34, Tim Connors wrote:
On Tue, 12 Mar 2013, Allan Duncan wrote:
It wouldn't be the first time a device has lied about its current draw or an adapter has lied about its ability to supply an amount of power.
You aren't kidding! I've got an Hitachi 1TB USB3 plugged in to a USB2 port, and lsusb reports: idVendor 0x4971 SimpleTech idProduct 0x1014 bcdDevice 0.00 iManufacturer 10 HitachiGST iProduct 11 Touro Mobile 3.0 ... MaxPower 8mA ...
Ooh, impressive! Must use superconducting magnets for the motors and maglev bearings. How do you go about supplying the liquid helium though? Obviously your cryopump isn't USB powered. Did you have to hook up 3 phase power?
Don't forget the vacuum pump - air friction is significant. Might have trouble getting the heads to fly, though.

On 12/03/13 13:23, James Harper wrote:
On 12/03/13 12:06, James Harper wrote:
. something else
Kernel bug (not saying it is here, just that it's always a possibility).
The reason I haven't made that assumption is that the previous 500GB disks still work fine.
Any idea how to easily report on USB power usage?
You can purchase dedicated little things for exactly this purpose. They're basically an digital amp meter with a usb port on each side, that sits inbetween the device and the host. I saw one for sale yesterday while looking for something else.

On Wed, 13 Mar 2013, Toby Corkindale <toby.corkindale@strategicdata.com.au> wrote:
You can purchase dedicated little things for exactly this purpose. They're basically an digital amp meter with a usb port on each side, that sits inbetween the device and the host.
I saw one for sale yesterday while looking for something else.
Where did you see it? -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

On 13/03/13 14:24, Russell Coker wrote:
On Wed, 13 Mar 2013, Toby Corkindale <toby.corkindale@strategicdata.com.au> wrote:
You can purchase dedicated little things for exactly this purpose. They're basically an digital amp meter with a usb port on each side, that sits inbetween the device and the host.
I saw one for sale yesterday while looking for something else.
Where did you see it?
If I could remember, I'd have posted the link. Sorry. It was one a site selling hobbyist gear for building robots and general embedded-hardware electronics gear; but I didn't end up buying from wherever the store was as I wasn't actually after that at the time.

On 12 Mar 2013, at 01:06, James Harper <james.harper@bendigoit.com.au> wrote:
[ 7455.667194] sd 16:0:0:0: [sdj] Attached SCSI disk [ 7690.448638] EXT3-fs error (device sdd1): ext3_find_entry: reading directory #2 offset 0 [ 7690.448972] Buffer I/O error on device sdd1, logical block 0 [ 7690.448994] lost page write due to I/O error on sdd1 [ 7690.448999] EXT3-fs (sdd1): I/O error while writing superblock [ 7690.449056] EXT3-fs (sdd1): error: ext3_journal_start_sb: Detected aborted journal
I've had similar errors on USB2 disks in HP ML350 G6's This has always been fixed by either moving the usb cable to the rear usb ports or using a dual connector usb cable, the ones with additional usb plug for power. maybe trying one of these or similar would resolve the issue - not used them personally. http://www.ebay.co.uk/itm/LINDY-1m-USB-3-0-Dual-Power-Cable-2-x-Type-Type-Ma... thanks Andy

I've had similar errors on USB2 disks in HP ML350 G6's
This has always been fixed by either moving the usb cable to the rear usb ports or using a dual connector usb cable, the ones with additional usb plug for power.
Actually so have I on those machines. The drives would work briefly then fall back to USB1 speeds.
maybe trying one of these or similar would resolve the issue - not used them personally.
http://www.ebay.co.uk/itm/LINDY-1m-USB-3-0-Dual-Power-Cable-2-x- Type-Type-Male-/120988999684#vi-content
I'd forgotten all about those splitter cables... I wonder if I can parasite power from a USB2 port (I have a 4 port USB3 adapter with all 4 ports in use, although only one to a drive with no external power which is the one I'm having problems with) James

James Harper <james.harper@bendigoit.com.au> wrote:
I'd forgotten all about those splitter cables... I wonder if I can parasite power from a USB2 port (I have a 4 port USB3 adapter with all 4 ports in use, although only one to a drive with no external power which is the one I'm having problems with)
You could try a powered USB 3 hub, provided that one can be obtained at a reasonable price.

James Harper <james.harper@bendigoit.com.au> wrote:
I'd forgotten all about those splitter cables... I wonder if I can parasite power from a USB2 port (I have a 4 port USB3 adapter with all 4 ports in use, although only one to a drive with no external power which is the one I'm having problems with)
You could try a powered USB 3 hub, provided that one can be obtained at a reasonable price.
I've had that thought too... just concerned it's another thing to go wrong :) Still... local IT store website says they have them for $37 with an AC adapter. James

James Harper <james.harper@bendigoit.com.au> wrote:
I've had that thought too... just concerned it's another thing to go wrong :)
Yes, but probably not straight away, and that should help you to determine whether power is the underlying problem here, rather than a kernel bug for example.
Still... local IT store website says they have them for $37 with an AC adapter.
That might be worth it, especially if you envisage more USB 3 device purchases.

On 13/03/13 10:21, James Harper wrote:
James Harper <james.harper@bendigoit.com.au> wrote:
I've had that thought too... just concerned it's another thing to go wrong :)
Still... local IT store website says they have them for $37 with an AC adapter.
hi this one is listed at $32 http://www.centrecom.com.au/catalog/m43hub-mbeat-port-plus-port-with-switche... Steve

On 13/03/2013 12:16 PM, Steve Roylance wrote:
On 13/03/13 10:21, James Harper wrote:
James Harper <james.harper@bendigoit.com.au> wrote:
I've had that thought too... just concerned it's another thing to go wrong :)
Still... local IT store website says they have them for $37 with an AC adapter.
this one is listed at $32 http://www.centrecom.com.au/catalog/m43hub-mbeat-port-plus-port-with-switche...
I had a similar one to that from COTD, it was useless -- even got a refund from them. Safeway/Woolies have some Dolphin 4 port USB hubs that are powered and the power adapters give reasonable amperage (5v 2A). The packaging says "Provides 4 additional Super Speed USB ports to your USB 3.0 enabled computer!", but you use one port on your computer, so you are only adding 3 ports. The price has ranged (over time) from $17 to $24.95, but they were $20 the last time I shopped a couple of days ago. I prefer to use drives that are not bus powered where possible as I've found that much more reliable. Perhaps that "lying" drive posted earlier is not bus powered, if so, then the max 8mA might be correct, from the USB port. Cheers -- Kind Regards AndrewM Andrew McGlashan Broadband Solutions now including VoIP Current Land Line No: 03 9012 2102 Mobile: 04 2574 1827 Fax: 03 9012 2178 National No: 1300 85 3804 Affinity Vision Australia Pty Ltd http://affinityvision.com.au http://securemywireless.com.au http://adsl2choice.net.au In Case of Emergency -- http://affinityvision.com.au/ice.html

I prefer to use drives that are not bus powered where possible as I've found that much more reliable.
Anything I don't plug in and out regularly will be 3.5" and so externally powered. Anything portable I will be plugging regularly and want to mess around with as few plugs as possible. It's disappointing, stuff like this should "just work", but here we are. James
participants (10)
-
Allan Duncan
-
Andrew McGlashan
-
Andy Dean
-
Chris Samuel
-
James Harper
-
Jason White
-
Russell Coker
-
Steve Roylance
-
Tim Connors
-
Toby Corkindale