| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 228.1 | | MOVIES::WIDDOWSON | Rod | Thu Feb 20 1997 12:21 | 23 |
| So the commands are
$ BAckup [dir...]*.* disk:[toto]
and
$ backup [dir...]*.* tape:/save
???
In this case the "/SAVE" is instructing backup to do completely
different things. In the former case the customer is doing what is
effectively a file by file copy (and at that to a single directory) and
in the other he is creating one saveset.
If this is the case, the customer will see a different time again when
he does
$ backup [dir...]*.* disk:[toto]sauve/save
Which is more analogous to the tape command
/r
|
| 228.2 | Disk may be fragmented | KEIKI::WHITE | MIN(2�,FWIW) | Thu Feb 20 1997 15:05 | 12 |
|
As a test
1. $Init the disk
2. $Mount the disk
3. $set rms/extend=1000
4. $Backup ......
To at least have a crabapple to apple test
Bill
|
| 228.3 | | AUSS::GARSON | DECcharity Program Office | Thu Feb 20 1997 17:10 | 21 |
| re .1
> So the commands are
>
> $ BAckup [dir...]*.* disk:[toto]
>
> and
>
> $ backup [dir...]*.* tape:/save
No, .0 says that /SAVE is on the disk.
re .0
As .2 says, check for fragmentation on the output disk and the RMS
extend size is very important.
Check for other user load on the output disk.
Also check where the various devices are in terms of buses and
controllers.
|
| 228.4 | I'd do this first | CERN::HOBBS | Congrats to the Ignoble Peace Prize winner! (http://www.eecs.harvard.edu/ig_nobel) | Fri Feb 21 1997 02:30 | 23 |
| .1> 3. $set rms/extend=1000
What is the extendize? Please note that BACKUP uses the default, so unless
something has been set backup will
* go to write a single buffer
* notice that the file needs to be extended
* extend the file by the default (or maybe MAX (default,buffersize))
* access the bitmap
* modify the file header
* write the file header to the disk
* read the file header back from the disk (data-check)
* write the buffer
Not only are you doing lots of extra reads and writes, those extra reads
and writes are pulling the head away from the data being written.
And, I'd think that /extend=1000 is pretty mild - I usually use /ext=65535
for any significant backup (you are going to have to truncate the file
at the end of the backup anyway, so what's the big deal if you truncate
40,000 blocks or 400?)
-cw
|
| 228.5 | replied to quickly | MOVIES::WIDDOWSON | Rod | Fri Feb 21 1997 04:24 | 1 |
| ... my mistake....
|
| 228.6 | Alternative w/disk-to-disk | GREGOR::OPP | | Fri Feb 21 1997 07:12 | 17 |
| RE: .0
If you (the customer) is going to do a disk-to-disk back-up,
why not preserve the directory structure in the backup:
BACKUP /VERIFY SRC$:[source...]*.*;* DST$:[000000...]
It's been a while since I've actually done this and the syntax is
a bit tricky as I recall. But if you're backing up the disk to
another disk, it's not *necessary* to flatten the directory or
encapsulate all the files & directories in a save set.
However, I'm sure the customer has a good reason for doing it
this way. Right?
Greg
|
| 228.7 | Simple reason | PADKOA::COSTEUX | Le Plat Pays qui est le mien... | Fri Feb 21 1997 10:22 | 5 |
| As the tape was unavailable the customer decided to make a backup on a
disk and stated the problem mentioned in .0
Thanks all for your comments.
Jean-Pierre
|