I worked some with the Oracle gateway back in 1995. I think that was
one of the last releases. The first thing that comes to mind is that
indeed the change that was rolled in on the Oracle side is the culprit.
Proving it might be difficult. It would be useful if you provided a bit
more information about the system abort. Maybe the number and the
entire message could be posted.
I will go out on a limb here and ask you if you have updated statistics
on the Oracle side to make sure you could get the best throughput
possible for the transactions. A minute per transaction sounds awful
long to me, and could leave the calling processes hanging in mid-air for
a long time.
Denys
-----Original Message-----
From: HP-3000 Systems Discussion [mailto:[log in to unmask]]On
Behalf Of Emerson, Tom
Sent: Tuesday, November 11, 2003 8:23 PM
To: [log in to unmask]
Subject: [HP3000-L] More or less ON topic: oracle gateway for 3k
(problem)
I'm kind of coming in on the tail end of this myself, but it seems we
have a process on an Oracle database [if you can say "ON an oracle
database..."] that is accessing an Image database via Oracle's "gateway"
[which I've just learned today is "unsupported"]
this process has run fine for (years, months, whatever...) and "all of a
sudden" caused a system failure. That's right -- flat out crashed the
box. I'm "told" that "no changes" have occurred [but I'll bet you've
heard that song-and-dance before...] (*) and this "process" has worked
with at least twice as many records as the current input file; it has
also processed the input file "on another system" without error, but
crash dumps are pointing a finger at this "gateway" and something to do
with a memory leak or bounds violation [you can imagine my skepticism of
THAT -- yes, this can happen, but it shouldn't bring down the box...]
[unless for some UNGODLY reason Oracle decided they MUST use PM rather
than remain in "user space" as they say in Unix-land]
Does any of this sound familiar? Is there a particular version of the
"gateway" that I should be looking for? [or at least to pass along to
the sysadmin to check against]
I've also learned that this system is excruciatingly slow -- it is
taking upwards of a minute PER TRANSACTION [and on a 4-way box to boot!]
so I'm suspecting the "query" that is being used is, shall we say, "less
than optimal"? [can you say, "serial read of the entire dataset..."? I
thought you could...] Is there any way I can prove [or disprove] this
theory? [for instance, where would I look for a "log" file showing what
transactions are occurring? I'm thinking something like the "odbclog"
that is written by the ODBC listener for normal "allbase/sql/PC"
traffic...]
Tom
(*) OK, confirmed that "no changes" was indeed a bit of a fib: the
process on the Oracle side of things has had a change rolled in on the
day this problem started. Not enough to point the finger solidly yet,
because if the problem is inherent in the gateway, it was just "waiting
to happen" and might have happened regardless of changes to the oracle
process...
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|
