From: Jon H. <jd_...@ya...> - 2009-01-06 15:57:08
|
I have a Model Family: Seagate Barracuda 7200.7 and 7200.7 Plus family Device Model: ST380011A and I have just noticed - only after i read somewhere about it - that my 'Seek_Error_Rate' keeps increasing? I have also now noticed that the 'Raw_Read_Error_Rate' and the 'Hardware_ECC_Recovered' are also increasing... What does this mean? I did notice that that Reallocated_Sector_Ct was set to 2.. I do short test daily and a long test 1 a week and it reports nothing... help! ----------------------- N: Jon Hardcastle E: Jon@eHardcastle.com '..Be fearful when others are greedy, and be greedy when others are fearful.' ----------------------- |
From: Michal S. <so...@zi...> - 2009-01-07 00:39:46
|
Jon Hardcastle wrote: > I have a Model Family: Seagate Barracuda 7200.7 and 7200.7 Plus > family Device Model: ST380011A > > and I have just noticed - only after i read somewhere about it - that > my 'Seek_Error_Rate' keeps increasing? I have also now noticed that > the 'Raw_Read_Error_Rate' and the 'Hardware_ECC_Recovered' are also > increasing... > This is normal with seagate drives. They are just extra verbose on those attributes. > What does this mean? I did notice that that Reallocated_Sector_Ct was > set to 2.. I do short test daily and a long test 1 a week and it > reports nothing... > Drive remapped 2 sectors so far. 7200.7 are pretty old drives (2003ish ?), so 2 sectors remapped during whole that time is probably nothing to really worry about. Depending on perspective and context, maybe even a reason to be proud of the drive ;) Either way, if you have important data, remember about backups (whenever the drive is 5 years old, or 5 days old). |
From: Jon H. <jd_...@ya...> - 2009-01-07 09:32:49
|
--- On Wed, 7/1/09, Michal Soltys <so...@zi...> wrote: > From: Michal Soltys <so...@zi...> > Subject: Re: [smartmontools-support] Should I worry? > To: Jon@eHardcastle.com > Cc: sma...@li... > Date: Wednesday, 7 January, 2009, 12:22 AM > Jon Hardcastle wrote: > > I have a Model Family: Seagate Barracuda 7200.7 > and 7200.7 Plus > > family Device Model: ST380011A > > > > and I have just noticed - only after i read somewhere > about it - that > > my 'Seek_Error_Rate' keeps increasing? I have > also now noticed that > > the 'Raw_Read_Error_Rate' and the > 'Hardware_ECC_Recovered' are also > > increasing... > > > > This is normal with seagate drives. They are just extra > verbose on those attributes. I have looked at my machine at home which sports 8 drives, 6 of which (mainly Hitachi) also seem to be uber verbose in these fields.. but the 'VALUE' seems to be pretty stable for these fields. Also if the THRESH is 0 what does that mean? there isn't one or that 0 is it? Is there anyway to find out which drives tend to be verbose? > > > What does this mean? I did notice that that > Reallocated_Sector_Ct was > > set to 2.. I do short test daily and a long test 1 a > week and it > > reports nothing... > > > > Drive remapped 2 sectors so far. 7200.7 are pretty old > drives (2003ish ?), so 2 sectors remapped during whole that > time is probably nothing to really worry about. Depending on > perspective and context, maybe even a reason to be proud of > the drive ;) Either way, if you have important data, > remember about backups (whenever the drive is 5 years old, > or 5 days old). This drive is old, it has done nearly 4 yrs of 'power on service' and i was horrified when i saw they were using it. Although I am now confident the drive is ok, I wouldn't want a drive that old at home - let alone carrying work critical data! I checked this field at home, and of my 8 drives most are 0 for this field but i notice that 2 hitachi's have a raw value of 102 and 104 and are sde and sdf respectively. Is this a coincidence? or is there something about where these drives are located in my machine that is causing them to relocate more sectors? should i worry about this? Smart itself is not complaining as a result of the tests i do daily on all my drives. On a side note, is it true that badblocks will never come back with bad sectors as the HDD always silently relocates them? |
From: Michal S. <so...@zi...> - 2009-01-07 22:21:55
|
Jon Hardcastle wrote: > I have looked at my machine at home which sports 8 drives, 6 of which > (mainly Hitachi) also seem to be uber verbose in these fields.. but > the 'VALUE' seems to be pretty stable for these fields. Also if the > THRESH is 0 what does that mean? there isn't one or that 0 is it? Is > there anyway to find out which drives tend to be verbose? > I don't know about Hitachi - I have only two such drives (old models though), both have 0 in seek/read error fields. So it might be worth digging further if you should or shouldn't worry. value/worst/thresh are so called normalized values. The lower value - the worse condition. Passing below threshold suggests that the drive is in bad condition / could fail soon (from manufacturer/firmware perspective). For some attributes it's better to follow raw values (uncorrectables, relocations, pending, etc.), for the others - normalized ones. Either way, normalized value falling under threshold, is definitely a reason for immediate backup and RMA. > This drive is old, it has done nearly 4 yrs of 'power on service' and > i was horrified when i saw they were using it. Although I am now > confident the drive is ok, I wouldn't want a drive that old at home - > let alone carrying work critical data! 4 years old drive that has been and is working well is often "safer" than brand new drive. YMMV. > I checked this field at home, and of my 8 drives most are 0 for this > field but i notice that 2 hitachi's have a raw value of 102 and 104 > and are sde and sdf respectively. Is this a coincidence? or is there > something about where these drives are located in my machine that is > causing them to relocate more sectors? should i worry about this? > At some point in time they developed bad or "unstable" sectors and reallocated them during writes. If the drives don't get new bad sectors or show up any alarming behaviour / problems, you're probably safe. Still - 102 / 104 is quite a few. Be sure to keep an eye on those hdds. > On a side note, is it true that badblocks will never come back with > bad sectors as the HDD always silently relocates them? If there're problems with reading sectors, the drive will mark them as "pending", and might reallocate them at next write, and depending on firmware and enabled features - on successful (but problematic) read as well. |
From: Jon H. <jd_...@ya...> - 2009-01-08 09:44:31
|
--- On Wed, 7/1/09, Michal Soltys <so...@zi...> wrote: > From: Michal Soltys <so...@zi...> > Subject: Re: [smartmontools-support] Should I worry? > To: Jon@eHardcastle.com > Cc: sma...@li... > Date: Wednesday, 7 January, 2009, 10:21 PM > Jon Hardcastle wrote: > > > I have looked at my machine at home which sports 8 > drives, 6 of which > > (mainly Hitachi) also seem to be uber verbose in these > fields.. but the 'VALUE' seems to be pretty stable > for these fields. Also if the > > THRESH is 0 what does that mean? there isn't one > or that 0 is it? Is there anyway to find out which drives > tend to be verbose? > > > > I don't know about Hitachi - I have only two such > drives (old models > though), both have 0 in seek/read error fields. So it might > be worth digging further if you should or shouldn't > worry. > > value/worst/thresh are so called normalized values. The > lower value - the worse condition. Passing below threshold > suggests that the drive is in bad condition / could fail > soon (from manufacturer/firmware perspective). For some > attributes it's better to follow raw values > (uncorrectables, relocations, pending, etc.), for the others > - normalized ones. > > Either way, normalized value falling under threshold, is > definitely a reason for immediate backup and RMA. > > > This drive is old, it has done nearly 4 yrs of > 'power on service' and > > i was horrified when i saw they were using it. > Although I am now confident the drive is ok, I wouldn't > want a drive that old at home - > > let alone carrying work critical data! > > 4 years old drive that has been and is working well is > often "safer" than brand new drive. YMMV. > > > I checked this field at home, and of my 8 drives most > are 0 for this > > field but i notice that 2 hitachi's have a raw > value of 102 and 104 and are sde and sdf respectively. Is > this a coincidence? or is there > > something about where these drives are located in my > machine that is causing them to relocate more sectors? > should i worry about this? > > > > At some point in time they developed bad or > "unstable" sectors and reallocated them during > writes. If the drives don't get new bad sectors or show > up any alarming behaviour / problems, you're probably > safe. Still - 102 / 104 is quite a few. Be sure to keep an > eye on those hdds. > > > On a side note, is it true that badblocks will never > come back with > > bad sectors as the HDD always silently relocates them? > > If there're problems with reading sectors, the drive > will mark them as "pending", and might reallocate > them at next write, and depending on firmware and enabled > features - on successful (but problematic) read as well. Thanks for your input I have taken this all on board - particularly about the old drive working well being better than an unknown quantity. Could you (or anyone else reading this) advise on a the best way to stimulate a read/write check to flush these out? I have 6 drives all raided together with LVM sat on top. I recently did a efsck of one of these LV's and did -cc i think which does a non-destructive read/write check. But this presumably didn't check all the disks as that LV will not span across all 6 drives in their entirety. Is it best to dissemble the array and do a badblocks on the drives individually? I did try a check/repair but wasn't sure if this checked every block on all the drives... Also, on that note. I'd like to did a read/write check on the blade at work.. BUT it has only 2 ports and both have the hard drives needed for the running of the machine. I have configured it using tune2fs to run a check on reboot but that wont (i believe) be a badblocks (read/write or even just read) but just a file consistancy check... so how might i do such a check on a mount file system? Thank you :) |
From: Martin B. <mar...@ic...> - 2009-01-08 11:23:14
|
Hi Jon, > Could you (or anyone else reading this) advise on a the best > way to stimulate a read/write check to flush these out? I > have 6 drives all raided together with LVM sat on top. I > recently did a efsck of one of these LV's and did -cc i think > which does a non-destructive read/write check. But this > presumably didn't check all the disks as that LV will not > span across all 6 drives in their entirety. Is it best to > dissemble the array and do a badblocks on the drives > individually? I did try a check/repair but wasn't sure if > this checked every block on all the drives... Depends on the raid you're using - in case of linux software raid a media check/repair facility is built into recent kernels; this requires raid levels with redundancy (i.e raid 1/5/6). echo "repair" > /sys/block/md0/md/sync_action (replace md0 with the raid device you want to check). This will * read all physical devices/partitions making up the device * rewrite a sector with read errors to allow drive sector remapping. If you just want problems to be logged without the rewurte, you can use "check" instead of "repair"; Bye, Martin |
From: Jon H. <jd_...@ya...> - 2009-01-08 14:13:37
|
--- On Thu, 8/1/09, Martin Bene <mar...@ic...> wrote: > From: Martin Bene <mar...@ic...> > Subject: AW: [smartmontools-support] Should I worry? > To: "'Jon@eHardcastle.com'" <Jon@eHardcastle.com>, "Michal Soltys" <so...@zi...> > Cc: "sma...@li..." <sma...@li...> > Date: Thursday, 8 January, 2009, 10:43 AM > Hi Jon, > > Could you (or anyone else reading this) advise on a > the best > > way to stimulate a read/write check to flush these > out? I > > have 6 drives all raided together with LVM sat on top. > I > > recently did a efsck of one of these LV's and did > -cc i think > > which does a non-destructive read/write check. But > this > > presumably didn't check all the disks as that LV > will not > > span across all 6 drives in their entirety. Is it best > to > > dissemble the array and do a badblocks on the drives > > individually? I did try a check/repair but wasn't > sure if > > this checked every block on all the drives... > > Depends on the raid you're using - in case of linux > software raid a media check/repair facility is built into > recent kernels; this requires raid levels with redundancy > (i.e raid 1/5/6). > > echo "repair" > /sys/block/md0/md/sync_action > > (replace md0 with the raid device you want to check). This > will > * read all physical devices/partitions making up the device > * rewrite a sector with read errors to allow drive sector > remapping. > > If you just want problems to be logged without the rewurte, > you can use "check" instead of "repair"; > > Bye, Martin Ah ok, Thanks martin. I did a check last night and the number of pending sectors didn't increase. I left a 'repair' going (even tho the check found nothing) hoping to coax them out. Will a repair write to each sector even if it didn't fail to read? Reading it may not cause problems whereas a write will.. |
From: Christian S. <sk...@ar...> - 2009-01-07 20:59:55
|
Hi Jon, > I checked this field at home, and of my 8 drives most are 0 for this > field but i notice that 2 hitachi's have a raw value of 102 and 104 and > are sde and sdf respectively. Is this a coincidence? or is there > something about where these drives are located in my machine that is > causing them to relocate more sectors? should i worry about this? Maybe the drives are located in a hotspot or have a bad mounting. Other than that, it might just be coincidence. A relocated sector count of 102 / 104 sounds a little bit high to me. I'd be cautious with these drives. Backups and maybe shortened sel-test intervals won't be too bad. On the other hand: if these are the only drives you have on this series, it might be that they use this Value in a different way. > Smart itself is not complaining as a result of the tests i do daily on > all my drives. The Tests will only fail on errors - e.G. unreadable sectors or something like that. Depending on the way you run smartd or smartctl, you will also be alerted if a normalized value dips below "Threshold". > On a side note, is it true that badblocks will never come back with bad > sectors as the HDD always silently relocates them? The HD will silently remap the sector if writing fails. If reading fails, the sector will not be remapped. The HD will mark this Sector as pending instead. RAW-Value #197 "Current_Pending_Sector" indicates the count of pending sectors. If a subsequent write to this area succeeds, the drive won't remap it. After all the reading error could have resulted from a power failure during writing or similar. Also note: every drive can only remap a limited amount of blocks. If it runs out of space in the specially designated area, the bad blocks will stay. Chris. |