Hello Mike
Back in 1991, we dd'd SEGY tapes on 9 track and used very large block sizes.
SEGY format consists of a 3200 byte record and then the subsequent record sizes
depend on the sampling rate and length of trace. We then analysed the tape
with tcopy to verify the record sizes and dd had indeed not pad the records, we
also read them with seismic applications on the MVS mainframe which would not
tolerate padded records and that worked fine as well. That was a long time
ago, the platform was SunOS for the dd but back then I can definitively say dd
did not pad the records. Now I believe there was an option to pad the records,
could be that the default behaviour on your platform was to pad.
Nothing like empirical evidence...
Regards,
John I Wang
Sr. Systems Engineer
Steverson Information Professionals
---
Enron Broadband Services
Enron Building 1472c
ph (713) 345-4291
fax (713) 646-8063
|--------+----------------------->
| | wei@colltech.|
| | com |
| | |
| | 10/13/00 |
| | 06:16 PM |
| | |
|--------+----------------------->
>----------------------------------------------------------------------------|
| |
| To: John Wang/Contractor/Enron Communications@Enron |
| Communications |
| cc: wei AT colltech DOT com, david AT datastaff DOT com,
|
| Devon_Buffington AT aimfunds DOT com, veritas-bu AT
mailman.eng.auburn DOT edu |
| Subject: Re: [Veritas-bu] Duplicating tapes |
>----------------------------------------------------------------------------|
Actually, dd won't do a "faithful" copy if you use a larger block factor to
read a tape and write to another tape. Say if dd is reading a tape record
of 8k in size and you use the block factor of 32k, dd will read those 8k,
pad it with 24k of zeros and write out as a 32k block to the duplicate tape.
The corresponding record on the duplicate tape will be 32k in size.
However, the duplicated tape *might* be usable by the program that will use
it. When the program reads this particular record, it might do a single
record read and only takes the first 8k worth of data. The way of Unix tape
device driver works is that the next tape read will be starting from the
next record. The remaining 24k of this record will be discarded.
Whether such duplicated tape will be usable depends on the behavior of the
program that uses it.
--Mike
> From veritas-bu-admin AT mailman.eng.auburn DOT edu Fri Oct 13 17:14:32 2000
>
> Hello David
>
> I used to come across variable block size problems all the time with various
> Seismic formats, what I found was that if you simply dd with a block size
larger
> than the largest that you'd expect to encounter on the tape, dd will produce a
> faithful copy but will complain about copying over nothing but partial blocks.
> So long as you're not pedantic about the block count message that dd will
print
> on the screen, it will work and be relatively painless.
>
> Yes, I too went through a period of time when I kept doing mt bsf to back over
a
> end of file marker so that I could do each record with an appropriately
> formulated dd command but then one day I said "What the hell..." and typed in
a
> huge block factor and was pleasantly surprised.
>
> Regards,
> John I Wang
> Sr. Systems Engineer
> Steverson Information Professionals
>
> ---
> Enron Broadband Services
> 3 Allen Center, Room 337C
> PH (713) 345-6238
>
>
>
>
> |--------+----------------------->
> | | wei@colltech.|
> | | com |
> | | |
> | | 10/09/00 |
> | | 02:59 PM |
> | | |
> |--------+----------------------->
>
>----------------------------------------------------------------------------|
> |
|
> | To: david AT datastaff DOT com
|
> | cc: Devon_Buffington AT aimfunds DOT com,
|
> | veritas-bu AT mailman.eng.auburn DOT edu, (bcc: John
> Wang/Contractor/Enron
|
> | Communications)
|
> | Subject: Re: [Veritas-bu] Duplicating tapes
|
>
>----------------------------------------------------------------------------|
>
>
>
>
> The problem with dd'ing Netbackup tapes is the block factor. You have to
> use the corrct block factor when you dd to make an useable tape. For
> non-NDMP tapes, it's fairly simple 'cause the first tape file is 2 blocks
> (1k) which is the tape lable, and the remaining tape files are all with
> block factor of 64 (32k block) if your backup server is on Solaris.
>
> For NDMP tapes however, the NDMP tape file has variable record size.
> I remember the first *record* of a NDMP backup image (tape file) is 2k (I
> could be a bit off here), and the remaining of the NDMP backup image has the
> block factor of 126 (63k record size). Getting this right while dd'ing
> isn't that straigtforward.
>
> dd'ing Netbackup tape isn't as easy as it seems.
>
> ...
>
>
>
>
> _______________________________________________
> Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
>
--
Mike Wei Collective Technologies, a Pencom Company
Email: wei AT colltech DOT com
|
