NEWS: Android Outsells Apple iPhone
in [Wireless Internet]
|
Prev: Introducing Poken Take your Social Network Offline
Next: Desperate Help Needed!! WIRELESS SHARING
From: Jeff Liebermann on 19 May 2010 20:53 On Wed, 19 May 2010 18:46:59 -0400, "Richard B. Gilbert" <rgilbert88(a)comcast.net> wrote: >Whether or not a backup can be "done with the operating system" or not >depends on the operating system. Backing up files while they are open >for write is usually just asking for problems! Well, one problem with image backups done with the OS running is that it's impossible to run a verify comparison with the original. I do this for my automated backups where it sometimes identifies the beginnings of read errors. Without the verify, I'm always hoping (praying) that I have a valid image backup. In this case, I lost. >The safe and sure way to make a backup is to take the file system >off-line. You have to force the O/S to commit ALL transactions to the disk. VSS allegedly does exactly that. There are other schemes that also work. However, after my last disaster, I'm not taking any chances. I boot from a CD and have absolutely no files open. >There are schemes to get around this but I'm not sure how well they work. Well, I know of one that I mentioned (Farstone Drive Image 5.0) that certainly does NOT work. Shadow Protect uses VSS and works quite well. However, I'm still paranoid: <http://www.storagecraft.com/shadow_protect_desktop.php> <http://www.storagecraft.com/kb/questions.php?questionid=102> >I'd suggest that five year old disks should not be trusted with anything >important. Five years is where the right hand side of the "bathtub >curve" climbs steeply! I've done several preemptive drive replacements after a few years. Using various S.M.A.R.T. utilities have helped predict failure, but my results are inconsistent. I'm still getting early drive failures well before 5 years. With an image backup and restore, the entire process can be done in about 3 hours (1.5 hrs billable) plus the cost of the new drive. The only trick is that I like to burn in my new drives to avoid infant mortality failures (which are rare, but very embarrassing). -- Jeff Liebermann jeffl(a)cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
From: Richard B. Gilbert on 19 May 2010 21:26 Jeff Liebermann wrote: > On Wed, 19 May 2010 18:46:59 -0400, "Richard B. Gilbert" > <rgilbert88(a)comcast.net> wrote: > >> Whether or not a backup can be "done with the operating system" or not >> depends on the operating system. Backing up files while they are open >> for write is usually just asking for problems! > > Well, one problem with image backups done with the OS running is that > it's impossible to run a verify comparison with the original. I do > this for my automated backups where it sometimes identifies the > beginnings of read errors. Without the verify, I'm always hoping > (praying) that I have a valid image backup. In this case, I lost. > >> The safe and sure way to make a backup is to take the file system >> off-line. You have to force the O/S to commit ALL transactions to the disk. > > VSS allegedly does exactly that. There are other schemes that also > work. However, after my last disaster, I'm not taking any chances. I > boot from a CD and have absolutely no files open. > >> There are schemes to get around this but I'm not sure how well they work. > > Well, I know of one that I mentioned (Farstone Drive Image 5.0) that > certainly does NOT work. Shadow Protect uses VSS and works quite > well. However, I'm still paranoid: > <http://www.storagecraft.com/shadow_protect_desktop.php> > <http://www.storagecraft.com/kb/questions.php?questionid=102> > >> I'd suggest that five year old disks should not be trusted with anything >> important. Five years is where the right hand side of the "bathtub >> curve" climbs steeply! > > I've done several preemptive drive replacements after a few years. > Using various S.M.A.R.T. utilities have helped predict failure, but my > results are inconsistent. I'm still getting early drive failures well The "bathtub curve" has two ends! If you plot failures vs. time, for a large sample, you get the so called "bathtub curve". It's called that because it's shape resembles a bathtub. \ / \ / \ / \ / \_________________________________/ The X axis represents time. The Y axis is number of failures. The left hand side shows "infant mortality" and the time scale is about 90 days. If something survives the first 90 days it will have to wait for accident or old age. The right hand side shows death from old age. For disk drives, the bottom of the tub represents two to six years. The early disk drives ran at maybe 3500 RPM. The RPM went up to 5400 and then 7200 RPM. Back around 2004 drives were introduced that spun at something like 15,000 RPM. Performance was great while the drives lasted. Infant mortality was terrible; we had a LOT of failures in the first 90 days. The boss decided that maybe 7200 RPM was the better choice. <snip>
From: John Navas on 20 May 2010 11:16 On Wed, 19 May 2010 17:53:57 -0700, Jeff Liebermann <jeffl(a)cruzio.com> wrote in <v619v5hhsdmb8n2avh5u0115bce6vgcn55(a)4ax.com>: >On Wed, 19 May 2010 18:46:59 -0400, "Richard B. Gilbert" ><rgilbert88(a)comcast.net> wrote: > >>Whether or not a backup can be "done with the operating system" or not >>depends on the operating system. Backing up files while they are open >>for write is usually just asking for problems! > >Well, one problem with image backups done with the OS running is that >it's impossible to run a verify comparison with the original. I do >this for my automated backups where it sometimes identifies the >beginnings of read errors. Without the verify, I'm always hoping >(praying) that I have a valid image backup. In this case, I lost. You're not matching your solution to your problem: 1. The most important verification is to check the readability of the backup, a good idea with WORM media (e.g., recordable DVD) but not of much value with external hard disk, which can (and should) be done without access to the original system. This should have prevented your loss unless the issue is media degradation, in which case verification wouldn't have helped. The best way to assure the viability of WORM media backup: (a) use premium media (like Verbatim media code MCC) and a good recorder -- all too many people use whatever's cheap and convenient. What do you use? ;) (b) test the burned media for a low soft error rate, which isn't done by most backup software -- a typical verify pass DOESN'T do that. Are you doing this? ;) (c) store the media properly -- an all too common problem. 2. The next most important verification is to be able to restore the backup, which can only (and should) be done in a synthetic test environment -- a typical verify pass DOESN'T do that. 3. The least important verification is to compare the backup to the original, which will only catch errors in the backup software and is of little real value if 1 and 2 are done. >>The safe and sure way to make a backup is to take the file system >>off-line. You have to force the O/S to commit ALL transactions to the disk. > >VSS allegedly does exactly that. There are other schemes that also >work. However, after my last disaster, I'm not taking any chances. I >boot from a CD and have absolutely no files open. You're not matching your solution to your problem. Witchcraft isn't the answer. ;) >>I'd suggest that five year old disks should not be trusted with anything >>important. Five years is where the right hand side of the "bathtub >>curve" climbs steeply! > >I've done several preemptive drive replacements after a few years. >Using various S.M.A.R.T. utilities have helped predict failure, but my >results are inconsistent. Studies show that SMART *isn't* a good predictor of problems, only a good check of current status. >I'm still getting early drive failures well >before 5 years. With an image backup and restore, the entire process >can be done in about 3 hours (1.5 hrs billable) plus the cost of the >new drive. The only trick is that I like to burn in my new drives to >avoid infant mortality failures (which are rare, but very >embarrassing). There are *big* differences in hard drive models, and the usual cause of failure (at any time) is a poor drive in the first place. This is why I won't go with any drive model until it has proven itself in the market for at least a year, why my personal notebook computer has the excellent Hitachi TravelStar 7K500 instead of the problematic Seagate equivalent, which was rushed to market several months earlier than the Hitachi, much less a second tier brand and model. All too many people use whatever's cheap and convenient. -- Best regards, John <http:/navasgroup.com> "If the only tool you have is a hammer, you will see every problem as a nail." -Abraham Maslow
From: John Navas on 20 May 2010 11:24
On Wed, 19 May 2010 21:26:47 -0400, "Richard B. Gilbert" <rgilbert88(a)comcast.net> wrote in <drydnRykcv1FEmnWnZ2dnUVZ_tKdnZ2d(a)giganews.com>: >Jeff Liebermann wrote: >> I've done several preemptive drive replacements after a few years. >> Using various S.M.A.R.T. utilities have helped predict failure, but my >> results are inconsistent. I'm still getting early drive failures well > >The "bathtub curve" has two ends! If you plot failures vs. time, for a >large sample, you get the so called "bathtub curve". It's called that >because it's shape resembles a bathtub. > >\ / > \ / > \ / > \ / > \_________________________________/ > >The X axis represents time. The Y axis is number of failures. The left >hand side shows "infant mortality" and the time scale is about 90 days. Conventional wisdom, but wrong, as conventional wisdom so often is. The actual probability of failure curve for good modern hard drives, based on real field data, is more like this (10 year scale): | | | | | | | |. .......................... | .................................... -- Best regards, John <http:/navasgroup.com> "Good judgment comes from experience, and a lot of that comes from bad judgment." -Will Rogers |