Re: transition to 7.0.16
Date: 21 Jul 1994 21:30:12 -0400
Message-ID: <30n7f4$i6o_at_search01.news.aol.com>
In article <CszK80.97G_at_uunet!aadt>, vrajan_at_aadt.com (Rajan,V) writes:
i>Hi,
i>What are the new features( if at all) in 7.0.16 ?? i>How painful was the transition from 7.0.15 -> 7.0.16 ?
i>Rajan
*-----------------
I would add to the previous response (only one I've seen via THIS interface; starting to think Delphi is looking better and better for USENET stuff despite the lack of a GUI USENET interface, but that's another story.) (Oh, and just because this is coming into the newsfeed via AOL, please don't automatically classify me as a mindless dork. Read the rest, and then you may classify me as a mindless dork, but objectively. :) )
The 7.0.15.--> 7.0.16 migration fixed a MAJOR, nightmarish bug in ..15
that made me crazy.
At least in the AT&T Global Information Systems (GIS) [formerly NCR] 3550
environment running NCR UNIX SVR4 rel 2.01, there was just a teensy
problem with import and export to tapes.
Teensy as in fatal; it seems that if your export exceeded a certain size or length, IMP would freak out and give you all sorts of errors including the deadly --0008 ( 'I don't know what YOU think this is, but it ain't an ORACLE export I'm looking at') IMP error. I had to do some arguing with some of the WWS people, extending over a period of I'd say 4 hours (75% of the time on the phone with them) to learn that the REAL bug has two parts.
First, if you're on 7.0.15x, you need to apply a patch I hadn't heard about that fixes part of the problem. They also recommended to me to get the VOLSIZE parameter out of my .par file (what a pain -- you need to set up to fake someone pressing RETURN at an interactive session in a batch job, unless you want to import a bazillion Gb database in a terminal session... yes, there's ways to implement that kludge.) At least as far as I know, you can't tell IMP or EXP in a .par file you're using <unlimited> which interactive sessions assume if you press RETURN. sigh A very helpful guy that helped me deal with the Bug From Hell - Part I internetted me the patch that I hand built into imp.
Oh, fine. now new errors, including the aforementioned -008. After seemingly ceaseless arguing over the phone, directed principally around "your tape is bad. pick another one.' and my 'that would not really help much, given it's too old!' and "oh, well, someone must have written to it without your knowing it" .. & etc, fortunately my favorite night operator called and said, Scott, you know, that tape thing you've been running, boy, we are sure getting a lot of errors on the console. Along the lines of the blocksize expected didn't match what was on the tape. Hmmm.
Based on the error messages, the expected and found and thrown away numbers added up to exactly 65536, one of our magic numbers. I modified the import par file to add a RECORDLENGTH of 65536, and, guess what? It worked just dandy, and we now had our system back. Now I specify that in both imp and exp parfiles and we've not seen that problem come back.
THIS error was fixed in .16.x. A NEW error was introduced in at least the NCR 3000 series builds.
You get a really wierd system error when you try to create a snapshot based on a view that's got UNION ALL in it, and perhaps some of the other set operators. There is no patch currently at least for us, but it is fixed in 7.1.4.x. The workaround we use is just to build it as a real table (create table AS blah blah blah).
Hope some or any of this helps. Write me if you want more details.
R. (Scott) Hunter shunter_at_cfotax.attmail.com Database Administrator
AT&T Tax Services Company, Financial Services Organization, AT&T Corporation Received on Fri Jul 22 1994 - 03:30:12 CEST