It's been too long since I had to recover from a disk failure. (This
one isn't especially critical, but I wouldn't mind losing as few data as
possible.) (With decade-old disks, you might think it'd happen more.)

I have a recent save set: BACKUP /IMAGE /RECORD, and multiple (daily,
normally) save sets: BACKUP /SINCE = BACKUP. I'm restoring the /IMAGE
save set now. After that, the (confusingly written) instructions
( https://vmssoftware.com/docs/VSI_SYS_MAN_UTIL_REF_VOL_I.pdf page 217 )
say:

> [...] Next, mount the disk as a file-structured volume and restore the
> incremental save sets in reverse chronological order. [...]

If the daily, incremental BACKUP jobs were all made with
/SINCE = BACKUP, and _without_ /RECORD, why wouldn't restoring the
_latest_ incremental save set be enough of the incrementals? What would
an earlier incremental have that a later one wouldn't?

PS: The following text in the PDF doc is not well formatted (I claim):

[...] The /
INCREMENTAL qualifier can [...]

I suspect that that's not the only such instance.

And, incidentally, the fellow who thought that every one of these
deserved an info message was wrong (I claim):

%BACKUP-I-SYMNOTFLW, <file_spec> is a SYMLINK and BACKUP will not follow
the link

On Monday, May 2, 2022 at 11:48:07 PM UTC-4, Steven Schweda wrote:
> And, incidentally, the fellow who thought that every one of these
> deserved an info message was wrong (I claim):
>
> %BACKUP-I-SYMNOTFLW, <file_spec> is a SYMLINK and BACKUP will not follow
> the link

As you surmise, restoring the most recent incremental will result in the disk having the state at the time of that incremental backup since the record pass was done only at the /IMAGE date. Other incremental savesets will contain files that were created and deleted between the /IMAGE and the most recent incremental pass. I suspect that the reasoning for the reverse order is to restore files saved in each incremental pass when /RECORD is turned on for those passes. Your use of the incremental is different.

Dan

> [...] Your use of the incremental is different.

Thanks. It's possible that when I devised this scheme (20-30 years
ago?), it all made sense. I thought that it still did, until I read the
current documentation.

I've seen descriptions like "Incremental = since last backup", and
"Differential = since last full backup", but not in VMS documentation.

As usual, better documentation might be better.

On Tuesday, May 3, 2022 at 8:54:55 AM UTC-4, Steven Schweda wrote:
> > [...] Your use of the incremental is different.
>
> Thanks. It's possible that when I devised this scheme (20-30 years
> ago?), it all made sense. I thought that it still did, until I read the
> current documentation.
>
> I've seen descriptions like "Incremental = since last backup", and
> "Differential = since last full backup", but not in VMS documentation.
>
> As usual, better documentation might be better.

Since you indicated that the record pass was done only on the image backup, the incremental and differential are the same.
If the record pass is done at each incremental pass, than the restore effort would be different.

Am 03.05.2022 um 05:27 schrieb Steven Schweda:
> It's been too long since I had to recover from a disk failure. (This
> one isn't especially critical, but I wouldn't mind losing as few data as
> possible.) (With decade-old disks, you might think it'd happen more.)
>
> I have a recent save set: BACKUP /IMAGE /RECORD, and multiple (daily,
> normally) save sets: BACKUP /SINCE = BACKUP. I'm restoring the /IMAGE
> save set now. After that, the (confusingly written) instructions
> ( https://vmssoftware.com/docs/VSI_SYS_MAN_UTIL_REF_VOL_I.pdf page 217 )
> say:
>
>> [...] Next, mount the disk as a file-structured volume and restore the
>> incremental save sets in reverse chronological order. [...]
>
> If the daily, incremental BACKUP jobs were all made with
> /SINCE = BACKUP, and _without_ /RECORD, why wouldn't restoring the
> _latest_ incremental save set be enough of the incrementals? What would
> an earlier incremental have that a later one wouldn't?
>
>
> PS: The following text in the PDF doc is not well formatted (I claim):
>
> [...] The /
> INCREMENTAL qualifier can [...]
>
> I suspect that that's not the only such instance.

Many years ago when backup emerged the explanation was:

If you restore incremental savesets forward then each different version
of a file has to be written to disk until you get the most recvent version

If you restore incremental savesets backwards then you write the first
occurence of a file to disk - all other versions in prior savesets can
be skipped.

At that time writing to disk was very costly.

> Since you indicated that the record pass was done only on the image
> backup, the incremental and differential are the same.

Huh? What, exactly, is the same as what, exactly? After a BACKUP
/IMAGE /RECORD, the first (partial) BACKUP /SINCE = BACKUP [/NORECORD]
could be thought of either way. After that, they're all "differential",
not "incremental".

Using the non-VMS definitions, I have multiple (daily) "differential"
backups. I need to tell BACKUP to do "/INCREMENTAL" to process any of
them correctly.

> Many years ago when backup emerged the explanation was: [...]

Ok. The (unhelpful) manual now says:

> 2. Restore any incremental save sets since the last full backup, in
> reverse chronological order, using the /INCREMENTAL qualifier.
>
> After you restore the save sets in this order, the output disk volume
> contains the same files it contained when the most recent incremental
> save operation was performed.

It also says:

> Note that BACKUP restores the volume correctly regardless of the order
> in which the incremental save sets are applied; using reverse
> chronological order is most efficient.

I saw no explanation of what makes it "most efficient", but I could
believe that going forward in time would cause some older files/versions
to be restored and then deleted.

> As usual, better documentation might be better.

Still true, I claim.