Slashdot Mirror
New Seagate Drives Have Real Difficulties With Linux
wtansill writes "Seagate's Free Agent series of drives are not intended to be compatible with the Open Source operating system Linux. The Inquirer reports on the problem: an unhelpful power saving mode. 'The problem is to do with the power-saving systems on Seagate's latest range of drives and the fact that it is shipped already formatted to NTFS. The NTFS is only a slight hurdle to Linux users who have a kernel with NTFS writing enabled or can work mkfs. But the "power saving" timer is a real bugger. It will shut the drive off after several minutes of inactivity and helpfully drop the USB connection. When the connection does come back it returns as USB1 which is apparently as useful as a chocolate teapot.' Via Engadget, though, there is a solution!
Actually, it's only incompatible with Open Sauce operating systems, so Linux should be fine.
I bought a Free Agent and I have not been happy with it. Sometimes it works and other times it doesn't. I went online to see what other users had experienced and read similar comments. A few people never had any problem with and liked it, but most had issues setting it up or getting it to run. I don't like this product.
Disconnecting hard drives is a big problem for external devices. So is power saving, and laptop use especially. I'll bet that Seagate will sell a "Mac-compatible" version fairly soon that voids this problem, and it'll be compatible with Linux.
But this is an amazingly foolish mistake on Seagate's part.
The Drive works, you just have to use sdparm to clear the idle flag so the drive won't spin down at all. But this is bad, its a deliberately defective product and I hope someone sues. Make that lots of people.
Judging from the huge numbers of comments on NewEgg (I'd guess that it was at least 20% of the comments) that the drive died within days or months, this Linux-unfriendly idle flag setting is really just a minor irritation.
On the other hand, since many of the failure comments blamed it on overheating, perhaps Linux users from regions with real penguins will be OK.
I could buy an argument as "there is a development bug, but we are fixing it soon and we are very sorry for this, but the faulty drives will be replaced".
There is no way in hell, I buy an argument like "Our drives are not supposed to work with Linux".
Either they hire complete idiots for their tech support, or this a sign of something really really bad smelling as the OOXML scandal or the SCO scandal.
Anyway, now I won't buy any more Seagate drives, at least not until Seagate has cleared this mess up.I bought two of these drives (500GB) a couple months ago. I tried that fix on one (turned off standby spin-down via sdparm), but ultimately the drive failed in about a week (possibly from heat, but I also needed to plug and unplug it when running as the power switch was not responding properly). And despite any five year warranties, who is going to send a failed drive with all your data off to who-knows-where? Years ago, back when drives cost $1000 for 1GB, I did that twice -- once the manufacturer sent my fixed drive back to a different person, and another time they sent it to an old address. There is another issue with the drives, which is that the tower part is not very solidly attached to the base, so it is wobbly (hard to believe, but the connection of the base to the tower drive section seemed very loose on the one I tried -- in general that whole two-part design seems questionable to me from a ruggedness standpoint). The power button is very confusing too -- it barely moves (maybe its capacitance based?) and does not always seem to work as I might expect it to (which may also have lead to the failure, when I pulled the plug on it). I returned the other one unopened. Someday I might put the first in an external enclosure and see if it works at all (some people online report success with that, although it entails physically breaking the case to get the drive out from what I read), but even if it does I will never trust it. I would recommend avoiding these drives for anyone based on the wobbly design alone. Despite the warranty and previously liking Seagate (before they bought Maxtor), I've moved back to Western Digital drives and others -- at least WD drives just sit there without potentially wobbling if you put them on a computer case with the slightest vibration. They definitely look cool in operation with the glowing stripe, but it seems this iteration put style way before function.
A 21st century issue: the irony of technologies of abundance in the hands of those still thinking in terms of scarcity.
Sounds like a dilemma to me.
Beware: In C++, your friends can see your privates!
...they are slow and OS dependent, either you loose oceans of space (FAT formatted drives) or you can't write to them from some OS'es (NTFS formatted drives) or a Mac just reformats the whole drive because it can't read it. ...and they are fast... many af them now have gigabit ethernet.
A NAS cost a little more and got all features you need without any of the problems... and you can get them almost as small as a external 3,5" drives.
And when combined with this story: http://www.channelregister.co.uk/2007/12/07/western_digital_drm_crippled_harddrive/
A kindly Reg reader tipped us off that the remote-access HDD won't share media files over network connections. Which is, as you can see here, the entire stinking point of it.
It's a scary world full of potentially unlicensed media. We're fortunate there's a hard drive vendor willing to step forward and do some indiscriminate policing for us.
From the WD site:
"Due to unverifiable media license authentication, the most common audio and video file types cannot be shared with different users using WD Anywhere Access."
WD's list of banned file types encompasses over 35 extensions. This includes AAC, MP3, AVI, DivX, WMV, and Quicktime files. And why not -- Windows TMP files too.
Looks like there's something going on to push Windows as the only OS, leaving Linux and the rest up a creek with no hard drives at this rate. This is very disturbing.Take Nobody's Word For It.
I have two FreeAgentDesktop 500G's and also had this problem. I found a solution on the web and adapted it slightly to be automatic. Create this script:
/sys/class/scsi_disk/*; do
#!/bin/sh
for i in
if [ "`cat "$i/device/model"`" = "FreeAgentDesktop" ]; then
if [ "`cat "$i/allow_restart"`" -eq 0 ]; then
echo 1 > "$i/allow_restart"
fi
fi
done
And put it into cron to run every 10 minutes (FreeAgentDesktops timeout is 15 minutes). I have it on ubuntu 7.04 but the only dependencies I recognise is to have kernel 2.6, sysfs and cron, which should not be an issue. I guess there is a nicer way to do this (e.g. script for dbus/hotplug), feel free to improve.
Looks like there's a fairly good solution at NSLU2-Linux. Sounds like it might handle the reattachment better.
That said, while I initially liked USB attached disks, I've later found the issues with lack of SMART and other features over USB to be a showstopper for any serious use (ie, anything beyond a replacement for burning DVD's for sneakernet transmission). I'm no longer particularly surprised when the level of 'working' of such devices is found to be relative.
I read the reviews at NewEgg before I bought this drive, and I was aware of the spin-down problem. I bought the 500gb model and use it as my MythTV store/tmp/work drive. I partitioned it /dev/sda and formatted it ext2. In over four months, I've had zero problems, and it gets used every day when MythTV records TV programs, flags the commercials, and prepares files to burn to DVD's. It gets warm, but not hot. Yes, I would feel better if it would spin-down when not in use, but so far, I wouldn't hesitate to buy another one. If someone is planning a class action lawsuit, I hope it's for graphics cards or wireless cards instead of FreeAgent drives.
Don't buy crap! If Seagate is only capable of spitting out this kind of crap, choose another vendor with a similar product.
Western Digital's 1TB My Book World Edition external hard drive has been crippled by DRM for your safety.
From the WD site:
"Due to unverifiable media license authentication, the most common audio and video file types cannot be shared with different users using WD Anywhere Access."
You have 20 seconds to comply
WD's list of banned file types encompasses over 35 extensions. This includes AAC, MP3, AVI, DivX, WMV, and Quicktime files. And why not Windows TMP files too.
http://www.channelregister.co.uk/2007/12/07/western_digital_drm_crippled_harddrive/
I've avoided buying Seagate drives since they started botching the SCSI interface back on the 150 MEGABYTE drives. The drives would accept selection while spinning up and loading the firmware from the media, then hang the bus until power was cycled. I have SCSI adapters with jumpers labeled "Seagate" that would hold off scanning the SCSI bus for a couple of minutes to let the Seagates become ready. No problem like that with any other drive manufacturer. This problem lasted at least through the 2 GByte 3.5" Barracuda, since I've tested HBAs against them and seen it.
It doesn't surprise me at all that they still have incompetent firmware programmers.
Simple solution: stop buying Seagate products and your problems will be fewer.
A solution to the FreeAgent spin-down problem was published on Ubuntu forums back in July 2007:
http://ubuntuforums.org/showthread.php?t=494673
It works for me very well. Importantly, it does not disable disk's power control. Instead, it auto restarts the disk whenever needed.
Er ..... any HP-badged one?
Je fume. Tu fumes. Nous fûmes!
How about a crontab entry that just writes something to the drive and syncs it* every so often?
(*) Under Linux in its default configuration, the file system is abstracted. All write operations are cached, and reads can be served from cache. Generally (this is an oversimplification) if sync is not issued deliberately, nothing is decached until shutdown, unless RAM starts getting dangerously low (it's too smart to do disk caching in swap space). This has the side-effect that on a box with plenty of RAM, a file can be created, modified, read and deleted without ever seeing oxide. It also means that certain things such as old versions of exim (which created masses of temporary files) and complex MySQL queries using temporary tables, seem to run blisteringly fast on Linux and slow to a crawl on Solaris (whose default setting is to decache between write and read operations, so that the read is served from disk and not cache.)
Je fume. Tu fumes. Nous fûmes!
I notice all the talk is about USB.
These drives are SATA drives and the FreeAgent drive my sister bought last month has an eSATA interface as well as USB (other models include the so-called FireWire interfaces as well.)
Why use USB with these devices at all, strangling your potential I/O bandwidth?
Given that it's a long timeout (15'), I'd guess that you simply haven't run afoul of it. Or possibly your distro has a patched kernel and allows longer for the drive to reconnect. The problem -- as far as I can piece it together -- is that a standard kernel.org kernel is not allowing the drive enough time to restart properly. A race condition ensues. The drive -- having sent a USB2 message, which got ignored because the host timed out -- thinks that the host computer isn't USB2 capable, and so reverts to USB1.
Je fume. Tu fumes. Nous fûmes!
I checked a lot of forums, and even PC users are having a lot of issues with the drive, not just Windows-based ones. I wouldn't even bother with a work-around with this drive. It has reliability and driver issues even in the realm where it was designed to function. Seagate appears to have designed a dud here. Western Digital has way more reliable drives. (Just don't install their software and you'll be fine).
The problem seems not to be the power saving, but the drop of the USB connection, which AFAIK violates all standards. It seesm to mean that the computer has to know the drive is there, and that it should ignore the obviously crashed USB connection and just asume the drive is still fine. Linux does the right thing and disconnects the drive. My guess is that on Windows, there is either a more optimistic driver (i.e. one that makes the customer happy and hides the problem) or these Seagates actually need their own, special driver.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
Wait what? You're blaming linux because Seagate made a new drive that breaks the USB spec?
I'm flaming you and telling you that you are stupid because you are blaming linux for following the spec.
I came across this problem months ago when I used 2 freeagent 750gb drives as part of a backup solution. I tried the above solution, but it didn't work correctly for me, I have a cron job that does an rsync backup and a human that swaps the drives every 2 days.
To be effective the above solution needs to be run as soon as you plug in the drive ie before the drive goes to sleep.
The way I got round it was to buy a cheap usbsata enclosures from ebay, cracked open the freeagent (which contains a normal sata drive) and installed the drive into el-cheapo ebay enclosures.
Problem solved with the added bonus that I can now have my human drive changer tell if the drive is in use and not change it if they see that flashing blue LED.
I used to think that Seagate had the edge on hard disk technology now I'm not so sure, anyone noticed how hot their drives run? even the 7200 rpm ones??
MacOS Classic adopted a different behaviour; the Mac designers removed the eject button from the floppy disk drive, making it impossible to eject a disk without the OS having a chance to unmount it first. I'm not quite sure how they dealt with network drives, however. UNIX was designed as a multi-user system, so only the system administrator would be able to add and remove disks (everyone else would be using a dumb terminal away from the computer) and since UNIX system administrators are meant to know what they are doing it they were expected to mount and unmount disk manually.
I am TheRaven on Soylent News
As usual with Linux, at the time slashdot picks the story up, the problem has
been fixed for some time (10 days ago in Linus' tree, in various test trees quite a bit longer):
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=f09e495df27d80ae77005ddb2e93df18ec24d04a
Windows has always used a Heisenburg unmount strategy (i.e. you don't know whether the drive is unmounted until you try accessing it).
Yes, and the 'doze also freezes the whole frickin' desktop for a painful few seconds any time a new CD is 'detected' until it is identified and explorer has a chance to fiddle with it. It's one of the really annoying things about the 'doze. I have looked long and hard for any way of disabling this auto-mount 'feature' but it seems to be a bug deeply planted in explorer. I suppose a whole manual 'mount' mechanism would have to replace it.
I did figure out what service to disable to keep fricking XP from diddling around and popping up a 'helpful' spam-dialogue any time a usb drive is detected.
I think you're prettifying the situation by calling an awful kludge a 'heisenberg' strategy.
Firewire or e-sata externals are much better. I have a cheap usb external that carps out if you try to push to much data to it also firewire 400 is faster then usb 2.0 with less cpu over head.
These are Seagate disks in USB enclosures. The problem here is with the behavior of the USB bridge chipset, NOT THE DISK.
I've later found the issues with lack of SMART
For what that's worth. The Google paper didn't find that SMART gave much warning before failure. And a former Seagate engineer (in alt.folklore.computer) said that they had found that competitors' drives were failing to log SMART errors, to make the numbers look better. He said that he had argued that Seagate should brag about showing honest numbers, but that marketing had won the argument and now he didn't believe any manfacturer's hard drive's SMART reports.
If you simply return the drive as defective, they'll shrug their shoulders and assume it was just that one disk. Tons of Windows users might not even have noticed.
The point of suing them is so there's no mistake -- every single drive is defective -- and so they don't assume they can simply give you a replacement drive and everything will be OK.
Don't thank God, thank a doctor!
"Give Vista forgiveness for allowing a virus to install a rootkit, Cancel Allow?"
?!?!?!
What I've been told is that some Seagate drives hold their own firmware on a few reserved sectors, which a low level wipe destroys. Regardless, the best solution seems to be; avoid Seagate.
Have gnu, will travel.
' = minutes
" = seconds
Je fume. Tu fumes. Nous fûmes!
The relevant difference is that USB is synchronous whereas Firewire is asynchronous. In terms of raw bitrates, USB is faster (480 Mbps vs. 400 MBps), but with a USB HDD you have to wait until the current block is completely transferred before you can request the next one; this makes it impossible to take full advantage of that raw capacity. With Firewire you can request blocks to be queued for transfer as the bus becomes available, meaning that you have less latency and higher overall bus saturation.
Moving from theory to practice, I have an external HDD enclosure with both USB 2.0 and Firewire 400 connectivity. Bulk data transfer is measurably faster when the enclosure is connected through Firewire. If you have a Firewire port available I would certainly recommend using it rather USB 2.0 for external bulk data storage.
"The state is that great fiction by which everyone tries to live at the expense of everyone else." - Bastiat
"For what that's worth."
Yep. Definitely for what it's worth. Still, it's important not to misread the google report; IIRC, while failures werent necessarily preceeded by SMART warnings, when SMART did warn there was a fair likelyhood of impending failure. Not enough to merit immediate replacement for google or someone else with massive redundancy (40% or something chance of failure within a short time period), it was definitely enough to merit migrating the disk to junk-disk for the average person.
Interesting.
I just bought one of these drives last week, and formated it ext3. I couldn't figure out why it always seamed to back up my data fine, but then the next morning (if left on) would always come back with a journal entry corrupt. forcing a unmount, and a fsck, then remount.
Wonder if my systems journal updates were too close to this timeout, so occasionally they just miss. Maybe a machine with lower utilization % would never have a problem.
Being used for nightly backup, if I use ext2 this probably won't cause a problem. And why use a journal for a file system that will only ever have 2-3 tar files on it anyway.
I guess I will return the drive regardless though, no reason to use a device with a known timing issue lurking.
I would agree. I think at this point we would be better off if we didn't try to come up with some far fetched hack and just started warning everyone to stay off the Seagates.
Which kind of sucks for me, I am in the market for a new server and was interested in the Seagate products because they have done very well in the past. But I can't afford to buy 5 drives for my server to find out that they sort of kind of mostly work some of the time. I'm well that past that era of crappy hardware support for Linux -- that's so RedHat 5.0.
Don't buy Seagate.
USB2 was so "obsolete" as soon as Firewire 400 was released. Oh...yeah, USB2 was released after
FW400...USB2 was obsolete upon release -- they should have gone with higher performance FW400. With the same hard disk years ago, I tried a speed test over 3 buses: ATA, USB2, FW400.
Performance for ATA & FW both topped out in the low 20's: ATA ~25MB/s, FW400: ~24MB/s. But USB2 -- topped out at 12MB/s. (USB1.1 was around 1.2MB/).
Anything I tried comparing FW400 & USB2 showed FW400 both faster and more reliable. Now FW800 is out and it does work noticeably faster than FW400.
USB2 is for "toys", not for system critical hardware. Maybe it is ok for talking to lower capacity USB devices, but for something close to a high-speed external and portable protocol, FW800 seems to do quite well.
Dunno about compared to ESata, one prob with FW800, is it seems to be faster than the hard disks I've
tested, so far, so I don't know its top speed or how it fares next to ESata, but USB?? I don't know why,
but it's 480Mb/s seems to run measurably slower than FW's 400Mb/s speed and, obviously, is no comparison compared to FW800.