From: Serguei M. <mi...@ci...> - 2006-09-02 16:45:24
|
Hello, Please, see subject. Is that possible? I'm asking this because when you are trying to access some data on the disk and kernel reports read errors, and smartmontools report about uncorrectable sectors, the data typically can not be recovered any more. Too late... Now, disk hardware has ECC bytes to recover from correctable read errors. smartmontools report the total number of such recoveries in Hardware_ECC_Recovered attribute but without any details, like the sector number and frequency of these recoveries in this sector, and/or a number of wrong and corrected bytes/bits in the sector, etc. In theory, some kind of statistics regarding such minor failures could give an idea about sectors which are still readable but may fail soon. This would greatly improve the chance to duplicate/copy/protect data, or one could force the disk firmware to reallocate data in this sector preventing its total loss. What do you think about this. Is it possible? For example, look at readcd tool. It has an option to check CDs against C2 read errors which are not fatal, they are corrected in the CD drive hardware. However, the number of these errors may say that it is time to copy that CD on a new one and save data for the future. Similar capability exists for DVD drives. However, I don't know any Linux utility which supports this functionality. Do you? Best regards, Serguei. |
From: Bruce A. <ba...@gr...> - 2006-09-02 19:51:54
|
I don't know of any linux tools that support this. The problem is that the detailed information about which sectors are in trouble, and the kind of trouble, is vendor-specific. And as far as I know, none of the vendors document this, although they make use of it in vendor-specific tools. Bruce On Sat, 2 Sep 2006, Serguei Miridonov wrote: > Hello, > > Please, see subject. Is that possible? > > I'm asking this because when you are trying to access some data > on the disk and kernel reports read errors, and smartmontools > report about uncorrectable sectors, the data typically can not > be recovered any more. Too late... > > Now, disk hardware has ECC bytes to recover from correctable > read errors. smartmontools report the total number of such > recoveries in Hardware_ECC_Recovered attribute but without any > details, like the sector number and frequency of these > recoveries in this sector, and/or a number of wrong and > corrected bytes/bits in the sector, etc. In theory, some kind > of statistics regarding such minor failures could give an idea > about sectors which are still readable but may fail soon. This > would greatly improve the chance to duplicate/copy/protect > data, or one could force the disk firmware to reallocate data > in this sector preventing its total loss. > > What do you think about this. Is it possible? > > For example, look at readcd tool. It has an option to check CDs > against C2 read errors which are not fatal, they are corrected > in the CD drive hardware. However, the number of these errors > may say that it is time to copy that CD on a new one and save > data for the future. > > Similar capability exists for DVD drives. However, I don't know > any Linux utility which supports this functionality. Do you? > > Best regards, > Serguei. > > ------------------------------------------------------------------------- > Using Tomcat but need to do more? Need to support web services, security? > Get stuff done quickly with pre-integrated technology to make your job easier > Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo > http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 > _______________________________________________ > Smartmontools-support mailing list > Sma...@li... > https://lists.sourceforge.net/lists/listinfo/smartmontools-support > |
From: Sergey S. <sha...@us...> - 2006-10-09 21:29:37
|
On Sat, Sep 02, 2006 at 02:51:49PM -0500, Bruce Allen wrote: >I don't know of any linux tools that support this. The problem is that=20 >the detailed information about which sectors are in trouble, and the kind= =20 >of trouble, is vendor-specific. And as far as I know, none of the vendors= =20 >document this, although they make use of it in vendor-specific tools. There is a DOS tool, MHDD -- http://hddguru.com/content/en/software/2005.10.02-MHDD/ -- that might help here. Its author claims that if the drive takes 'too much' time to execute a command (VERIFY SECTORS), then these sectors are in trouble. There's an option to 'fix' these sectors (this function is called 'Erase Waits'). --=20 Sergey Svishchev |
From: Serguei M. <mi...@ci...> - 2006-09-02 20:56:47
|
In fact, it means that some data loss is unavoidable before user will start to worry about disk replacement. BTW, I don't know any vendor specific tool, for Hitachi disks, for example, that would show minor errors and suggest a user to reallocate those sectors. They typically scan the disk and reallocate sectors only when error becomes uncorrectable. Or, am I wrong? What is sad that disk manufacturers rarely consider the disk defective when sectors can still be replaced during warranty period. They all suggest to run some kind of Fitness Test or Disk Repair tool, and only issue RMA when there is no replacement sectors available. My Hitachi DK23DA-30 started to show trouble during first year of use in a new notebook. At that time the disk was almost empty and I almost had no problem reallocating about 35 sectors. After that 2.5 years - no a single error. In May 2005 another sector was in trouble - reallocated. One week ago, at the and of August 2006 - again. Now I'm starting to think about a new disk... What is criteria for new disk behavior? Is it normal to detect errors during first year of operation? Should user insist on RMA in such a case? What are some general considerations? Also, regarding purchase of a new disk for notebook, is there something like a database with hard disk reliability data? Which brand is best? Thank you and best regards, Serguei. On Saturday 02 September 2006 12:51, Bruce Allen wrote: > I don't know of any linux tools that support this. The > problem is that the detailed information about which sectors > are in trouble, and the kind of trouble, is vendor-specific. > And as far as I know, none of the vendors document this, > although they make use of it in vendor-specific tools. > > Bruce > > On Sat, 2 Sep 2006, Serguei Miridonov wrote: > > Hello, > > > > Please, see subject. Is that possible? > > > > I'm asking this because when you are trying to access some > > data on the disk and kernel reports read errors, and > > smartmontools report about uncorrectable sectors, the data > > typically can not be recovered any more. Too late... > > > > Now, disk hardware has ECC bytes to recover from > > correctable read errors. smartmontools report the total > > number of such recoveries in Hardware_ECC_Recovered > > attribute but without any details, like the sector number > > and frequency of these recoveries in this sector, and/or a > > number of wrong and corrected bytes/bits in the sector, > > etc. In theory, some kind of statistics regarding such > > minor failures could give an idea about sectors which are > > still readable but may fail soon. This would greatly > > improve the chance to duplicate/copy/protect data, or one > > could force the disk firmware to reallocate data in this > > sector preventing its total loss. > > > > What do you think about this. Is it possible? > > > > For example, look at readcd tool. It has an option to > > check CDs against C2 read errors which are not fatal, they > > are corrected in the CD drive hardware. However, the > > number of these errors may say that it is time to copy > > that CD on a new one and save data for the future. > > > > Similar capability exists for DVD drives. However, I don't > > know any Linux utility which supports this functionality. > > Do you? > > > > Best regards, > > Serguei. |
From: Geoff K. <ge...@ge...> - 2006-09-04 08:38:51
|
Serguei Miridonov <mi...@ci...> writes: > In fact, it means that some data loss is unavoidable before > user will start to worry about disk replacement. BTW, I don't > know any vendor specific tool, for Hitachi disks, for example, > that would show minor errors and suggest a user to reallocate > those sectors. They typically scan the disk and reallocate > sectors only when error becomes uncorrectable. Or, am I wrong? I had a disk, I think it was one of the IBM DeskStar disks, which was consistently allocating replacement sectors without losing any data. Eventually (after about two years) it ran out of replacement sectors its SMART status went to 'fail', and I replaced it. I was running a long self-test on it once a week, and short tests daily, and had the offline routine enabled; I think this helped a lot, because when I got the drive none of this was happening and it had one bad sector. According to the documentation for a Hitachi 80GN laptop drive, it will: > When a read operation for a sector fails and is recovered at the > specific ERP step, the sector is reallocated automatically. A media > verification sequence may be run prior to the reallocation according > to the pre-defined conditions. This is somewhat vague, and was probably a bit vague even in the original Japanese, but it sounds like what it'll do is that if it detects a read problem (a lot of single-bit errors which are either recovered by the ECC coding or too many for ECC but re-reading the sector succeeds), it will try to re-write the sector and then if that doesn't seem to clear up the read error it'll reallocate the sector. Obviously this will only work if you're ensuring that everywhere on the disk gets read occasionally, for example by running a long self-test. |
From: Bruno W. I. <br...@wo...> - 2006-09-02 21:39:04
|
On Sat, Sep 02, 2006 at 13:56:34 -0700, Serguei Miridonov <mi...@ci...> wrote: > In fact, it means that some data loss is unavoidable before > user will start to worry about disk replacement. BTW, I don't > know any vendor specific tool, for Hitachi disks, for example, > that would show minor errors and suggest a user to reallocate > those sectors. They typically scan the disk and reallocate > sectors only when error becomes uncorrectable. Or, am I wrong? That is one reason you want to run RAID. |
From: Serguei M. <mi...@ci...> - 2006-09-02 21:41:34
|
On Saturday 02 September 2006 14:39, Bruno Wolff III wrote: > On Sat, Sep 02, 2006 at 13:56:34 -0700, > > Serguei Miridonov <mi...@ci...> wrote: > > In fact, it means that some data loss is unavoidable > > before user will start to worry about disk replacement. > > BTW, I don't know any vendor specific tool, for Hitachi > > disks, for example, that would show minor errors and > > suggest a user to reallocate those sectors. They typically > > scan the disk and reallocate sectors only when error > > becomes uncorrectable. Or, am I wrong? > > That is one reason you want to run RAID. On a notebook? |
From: Bruno W. I. <br...@wo...> - 2006-09-02 21:47:05
|
On Sat, Sep 02, 2006 at 14:41:28 -0700, Serguei Miridonov <mi...@ci...> wrote: > On Saturday 02 September 2006 14:39, Bruno Wolff III wrote: > > On Sat, Sep 02, 2006 at 13:56:34 -0700, > > > > Serguei Miridonov <mi...@ci...> wrote: > > > In fact, it means that some data loss is unavoidable > > > before user will start to worry about disk replacement. > > > BTW, I don't know any vendor specific tool, for Hitachi > > > disks, for example, that would show minor errors and > > > suggest a user to reallocate those sectors. They typically > > > scan the disk and reallocate sectors only when error > > > becomes uncorrectable. Or, am I wrong? > > > > That is one reason you want to run RAID. > > On a notebook? Notebooks have other ways to lose data that are probabably more of a problem than losing a couple of sectors worth of data without advanced notice. You should be backing up data from your notebook to a safer location regularly. |
From: Serguei M. <mi...@ci...> - 2006-09-02 22:04:45
|
On Saturday 02 September 2006 14:55, Bruno Wolff III wrote: > On Sat, Sep 02, 2006 at 14:41:28 -0700, > > Serguei Miridonov <mi...@ci...> wrote: > > On Saturday 02 September 2006 14:39, Bruno Wolff III wrote: > > > On Sat, Sep 02, 2006 at 13:56:34 -0700, > > > > > > Serguei Miridonov <mi...@ci...> wrote: > > > > In fact, it means that some data loss is unavoidable > > > > before user will start to worry about disk > > > > replacement. BTW, I don't know any vendor specific > > > > tool, for Hitachi disks, for example, that would show > > > > minor errors and suggest a user to reallocate those > > > > sectors. They typically scan the disk and reallocate > > > > sectors only when error becomes uncorrectable. Or, am > > > > I wrong? > > > > > > That is one reason you want to run RAID. > > > > On a notebook? > > Notebooks have other ways to lose data that are probabably > more of a problem than losing a couple of sectors worth of > data without advanced notice. You should be backing up data > from your notebook to a safer location regularly. Sure. I do this using unison. Nevertheless, I think that current state of technology should provide some means to alert user in advance of possible data loss. |
From: Volker K. <lis...@pa...> - 2006-09-02 23:32:16
|
> Sure. I do this using unison. Nevertheless, I think that > current state of technology should provide some means to alert > user in advance of possible data loss. I totally agree, unfortunately, hard disk manufacturers do not, and as it's the software hidden inside the disks that would need to provide the desired information, I don't see any chance at all of doing this while being stonewalled by disk manufacturers. Current state of affairs is that disk manufacturers don't even care about whether users are told about the disk failure values available via the SMART standard, if RedmondOS(TM) is anything to go by. Those users don't get told by Billy what the story is. Those Dozers installing smartmontools are too few to count for disk makers to take notice. Volker -- Volker Kuhlmann is list0570 with the domain in header http://volker.dnsalias.net/ Please do not CC list postings to me. |