On Mon, Nov 21, 2016 at 12:59 PM, Robert Moskowitz <rgm@xxxxxxxxxxxxxxx> wrote:
On 11/21/2016 09:20 AM, Matthew Miller wrote:
On Mon, Nov 21, 2016 at 08:21:01AM -0500, Robert Moskowitz wrote:
I have a 1TB drive that is getting I/O errors enough that while INot helpful _now_, but consider making sure you have backups once you
What is a good way to move the OS and all the data? I used
Clonezilla some many years ago, but that does not seem to be the way
to go when there are bad sections on the source drive?
start using the new system. Then, next time this happens, forget
worrying about transferring from a failing drive and just restore your
backup.
The data is backed up. In fact it was one of the rsync cron tasks that wedged the system while I was gone. The I/O did not seem to be in the data, but perhaps some temp or other storage needed.
Backups should periodically read the entire disk, so may encounter bad sectors in areas that are not often read in normal operation. Backups that take
too long to run are often the first indication of a failing disk.
The challenge is the OS and software configs. Actually the system is a ClearOS file server and there are all those config information scattered around by ClearOS. I back up the whole /var/flexshare, but there is a lot more to their setup and no 'clear' directions from them. Building a new server from scratch would be painful. Last year (when I had time while building up my contract work) I looked into building my own Samba 4 server, but dropped it. Anyway, I have yet to find directions on a good way to maintain a running backup of the OS for a straight rebuild.
After many experiences with having to reinstall the OS in order to install the backup client needed to restore user data, I make it a practice to keep a
base OS with current updates handy. Without that I end up reinstalling the OS and then doing multiple updates. If you do a few updates a week and
hit a problem then it usually isn't hard to figure out which package caused the issue, but on a system that has been in service for a few years with 100's updates, getting form a fresh install from the original media back to an up-to-date running system may not be easy. With systems that don't have restrictive licenses, you can just dd the base system to a fresh disk, change the system name and add the user(s), then install the backup client, restore the user data, and reinstall end-user applications.
ddrescue often recovers nearly all the files, so even if the resulting image isn't fully working you may be able to recover most of the configuration data.
If this is a desktop system with room for it, consider buying *two*
drives and mirroring them. That's not a backup (since errors and
mistakes get propagated instantly, and because it's in the same
physical location), but makes situations like this rare (at least, if
you replace as soon as you start seeing errors instead of waiting for a
double failure).
I could never figure out what mirroring failing sectors meant.
Not sure of the context, but the first sign of a failing disk is often a massive decrease in
performance which I'm told is due to multiple re-reads of a sector. When a sector shows
signs of failing, disk firmware copies the data to a spare sector. If the new disk doesn't
signs of failing, disk firmware copies the data to a spare sector. If the new disk doesn't
arrive quickly you then see disk failure with a code that translates to "no spare sectors left
on device".
George N. White III <aa056@xxxxxxxxxxxxxx>
Head of St. Margarets Bay, Nova Scotia
Head of St. Margarets Bay, Nova Scotia
_______________________________________________ users mailing list -- users@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to users-leave@xxxxxxxxxxxxxxxxxxxxxxx
