From: Scott L. <sl...@sl...> - 2006-06-30 16:07:12
|
I've got a couple PowerBooks with drive trouble On one of them (with a Toshiba MK6021GAS), smartctl seems to say it's fine. It says "PASSED", there are only 3 error log entries, and "long" self tests have completed successfully. But accesses to parts of the disks are horrendously slow. I picked out this error entry: Error 1 occurred at disk power-on lifetime: 8874 hours (369 days + 18 hours) When the command that caused the error occurred, the device was active or idle. After command completion occurred, registers were: ER ST SC SN CL CH DH -- -- -- -- -- -- -- 40 51 a6 41 03 15 40 Error: UNC 166 sectors at LBA = 0x00150341 = 1377089 Commands leading to the command that caused the error were: CR FR SC SN CL CH DH DC Powered_Up_Time Command/Feature_Name -- -- -- -- -- -- -- -- ---------------- -------------------- c8 00 aa 3d 03 15 40 00 21:07:10.785 READ DMA ca 00 80 25 19 94 40 00 21:07:10.784 WRITE DMA ef 02 00 00 00 00 00 02 21:07:10.784 SET FEATURES [Enable write cache] ef aa 00 00 00 00 00 02 21:07:10.784 SET FEATURES [Enable read look-ahead] ef 03 45 00 00 00 00 02 21:07:10.784 SET FEATURES [Set transfer mode] and, in single-user mode, looped to read a cache full at that location: while true; do \ dd if=/dev/disk0 of=/dev/null bs=512 count=4097 skip=1377089 \ 2>&1 | fgrep transferred done and got horrible results - it took 5-10 seconds most of the time to read that much data. Another run (don't remember what sector I'm reading here) is even worse: 2097152 bytes transferred in 22.492957 secs (93236 bytes/sec) 2097152 bytes transferred in 24.061615 secs (87158 bytes/sec) 2097152 bytes transferred in 31.856649 secs (65831 bytes/sec) 2097152 bytes transferred in 18.171225 secs (115411 bytes/sec) 2097152 bytes transferred in 21.267094 secs (98610 bytes/sec) 2097152 bytes transferred in 22.266301 secs (94185 bytes/sec) 2097152 bytes transferred in 36.449603 secs (57536 bytes/sec) 2097152 bytes transferred in 26.026536 secs (80577 bytes/sec) 2097152 bytes transferred in 109.511686 secs (19150 bytes/sec) ... That's a couple orders of magnitude under the spec's 154.3-29.0Mb/s (19000000 bytes/sec - 3600000 bytes/sec). It seems to go in streaks - sometimes I'll get a reasonable 0.2 seconds or so for a while. Is there some way I'm missing to force the disk to admit this area is bad? -- Scott Lamb <http://www.slamb.org/> |
From: Martin S. <Ma...@li...> - 2006-06-30 16:26:13
|
Am Freitag 30 Juni 2006 18:07 schrieb Scott Lamb: > I've got a couple PowerBooks with drive trouble On one of them (with a > Toshiba MK6021GAS), smartctl seems to say it's fine. It says "PASSED", > there are only 3 error log entries, and "long" self tests have > completed successfully. But accesses to parts of the disks are > horrendously slow. Hello Scott, I recommend to have a look at this: http://blog.philkern.de/archives/173-Load-Cycle-Fix-for-Toshiba-HDDs.html > Is there some way I'm missing to force the disk to admit this area is > bad? I don't know one. Regards, -- Martin 'Helios' Steigerwald - http://www.Lichtvoll.de GPG: 03B0 0D6C 0040 0710 4AFA B82F 991B EAAC A599 84C7 |
From: Martin S. <Ma...@li...> - 2006-06-30 16:33:42
|
Am Freitag 30 Juni 2006 18:07 schrieb Scott Lamb: > Error 1 occurred at disk power-on lifetime: 8874 hours (369 days + 18 > hours) When the command that caused the error occurred, the device was > active or idle. > > After command completion occurred, registers were: > ER ST SC SN CL CH DH > -- -- -- -- -- -- -- > 40 51 a6 41 03 15 40 Error: UNC 166 sectors at LBA = 0x00150341 = > 1377089 Hello again, look at man smartctl: "UNC (UNCorrectable): data is uncorrectable. This refers to data which has been read from the disk, but for which the Error Checking and Correction (ECC) codes are inconsistent. In effect, this means that the data can not be read." Usually a disk automatically maps out bad sectors, but when data cannot be read it might refuse to map those out in the hope that they can be read a later time. AFAIK badblocks can generate a list of bad blocks, that you can feed to fsck.ext3, but I never tried that. I am using XFS here. Regards, -- Martin 'Helios' Steigerwald - http://www.Lichtvoll.de GPG: 03B0 0D6C 0040 0710 4AFA B82F 991B EAAC A599 84C7 |
From: Scott L. <sl...@sl...> - 2006-06-30 17:20:34
|
Hi Martin, Thanks for the replies. As to the blog entry, I'm running OS X, and my Load_Cycle_Count seems normal. On Jun 30, 2006, at 9:33 AM, Martin Steigerwald wrote: > Usually a disk automatically maps out bad sectors, but when data > cannot be > read it might refuse to map those out in the hope that they can be > read a > later time. Well, it seems to be reading them now (very, very slowly) and is not remapping them. The reallocated sector count is 0. In fact, I first noticed problems long before it logged any errors at all. It would just go into this state where the machine would be almost totally unresponsive, loudly seeking about twice a second. > AFAIK badblocks can generate a list of bad blocks, that you can > feed to > fsck.ext3, but I never tried that. I am using XFS here. If OS X and HFS+ have an equivalent of "badblocks", I haven't found it. There wasn't even any SMART utility; I compiled smartmontools from CVS. -- Scott Lamb <http://www.slamb.org/> |
From: Scott L. <sl...@sl...> - 2006-06-30 21:39:14
|
On Jun 30, 2006, at 10:20 AM, Scott Lamb wrote: > Thanks for the replies. As to the blog entry, I'm running OS X, and > my Load_Cycle_Count seems normal. Whoah, I must have been looking at the wrong line. ID# ATTRIBUTE_NAME FLAG VALUE WORST THRESH TYPE UPDATED WHEN_FAILED RAW_VALUE ... 9 Power_On_Hours 0x0032 066 066 000 Old_age Always - 13886 ... 12 Power_Cycle_Count 0x0032 100 100 000 Old_age Always - 3628 ... 193 Load_Cycle_Count 0x0032 008 008 000 Old_age Always - 921529 It's a three-year-old drive, so on average it's gone through a load cycle once every 3*365*24*60/921529 ~= 2 minutes. That's not cool. My other failing drive (in a six-month-old PowerBook) is doing once a minute. If that 300,000 load cycles thing is really true, I'll have to figure out a way to fix this on OS X before I replace the drive. I don't want the new one going the same way...thanks for the pointer. Toshiba's specs don't actually say - just: MTTF (Power on hours): 300,000 Product Life: 5 years or 20,000 power ON hours -- Scott Lamb <http://www.slamb.org/> |