Re: File transfer
On 02/28/2015 07:07 AM, David Wright wrote:
I don't know how to "have a look around. If everything looks good,
proceed with using the computer." Not a clue.
I call that the "smoke test" -- boot, log in, start some terminals, run
some commands, start some applications, play some games, start the file
manager, browse the file system, browse the network, play some music and
videos, surf the web, log in via SSH from another machine, reboot, log
in again, etc., looking for malfunctions. It's an informal,
non-repeatable, "quick and dirty" spot failure test. Most imaging
problems will cause a smoke test failure, but passing doesn't prove
correctness.
Ideally, there would be a complete system regression test to verify
correctness, and each user would run that from their desktop. Please
let me know if you find one.
OTOH having generated an md5sum of the backup, why not just pipe a
repeat dd into md5sum and see if they match. If you must be thorough,
then wipe and restore the system and dd | md5sum again.
You're right that it's probably more correct to feed the dd input to
md5sum, rather than the output. I assume dd will throw an error message
if the copying fails. I use USB flash drives for system drives, they
are larger than my RAM, and therefore my kernel buffer cache doesn't
contain all the blocks after dd runs. So, I md5sum the output because
it's on a hard drive and takes less time.
Another small point; by "do a fresh install", do you mean repeat your
restoration? Or install afresh, ie from scratch?
"fresh install" = "install afresh" = "install from scratch"
If the restored image fails to work correctly, there are at least the
following possibilities:
1. Hardware failure.
2. System drive contained the flaws when imaged.
3. Imaging process failed.
4. Image restoration process failed.
If #1 is the case, the machine should fail the smoke test after wiping
and re-installing. If I fix and verify the hardware, restoring a good
image should give me a working system.
That leaves #2-4, which I assume were caused by operator error. So, I
try again (wipe, install, smoke test, image, restore, smoke test) and
pay more attention.
An unrelated question is how often do you do all this, and how do you
age your image. By "age" I mean how long do you treat it as a valid
image because your live system is evolving from the moment you start
reusing it after imaging it.
I take system drive images after installation, prior to deployment,
infrequently (3~6 months), prior to making big scary changes, after
making big scary changes, and whenever else I feel the need.
Aging isn't too big of a deal, because of Approx -- most or all of the
updates since imaging should be local, so restoring, updating, and
upgrading is fast.
Myself, I prefer to archive (original and modified) copies of any
system files I change, any configuration commands I've used, package
lists, non-Debian debs etc. and re-install from scratch.
I save downloads on the file server. I keep a system administration
journal, a package list, and any configuration files I modify under
version control (CVS). After re-install, I check out the old
configuration files to a scratch directory and do the edits manually.
(I've played with scripting certain common post-install chores, with
limited success.)
(We're going off topic...)
Thank you for bringing me to the point of my "backup" digression:
You transfer your files from an old machine to a new machine by
backing up your files on the old machine and restoring them on the
new machine.
David
Reply to: