Shell Programming and Scripting

Well I think post #6 is pretty much what methyl suggested.

Thanks Chubler for the code. I ran it a couple of times and I got the same issue. Hits a long running sql script and ignores the others.

This is the output I was getting:

NUM= 0
[1] 3278
NUM= 1
[2] 3282
NUM= 2
[3] 3286
NUM= 3
[4] 3290
NUM= 4
[5] 3294
NUM= 5
[6] 3298
NUM= 6
NUM= 6
NUM= 6
NUM= 6
NUM= 6
NUM= 6
..........

It gets stuck at 6 and doesn't proceed further. I check the sqlplus jobs and I find nothing is running:

ps -ef|grep "sqlplus -s"

oracle 4341 15634 0 20:08:09 pts/1 0:00 grep sqlplus -s

@gctex
Please post the complete and unedited code (after blotting confidential information with X's) which produced the output in post #9.

Ps. I suspect that I first became a certified Oracle DBA before you were born. There is much popular usage of basic computer terms in general circulation. In the context of your posts it is impossible to determine whether you are referring to a program written in Oracle SQL or a unix Script. In my above request I use the word "code" to refer to either.

@gctex, From the output (I left the NUM=$NUM debugging line in by mistake), we can see it has fired off 6 jobs and the fact that NUM=6 is poping up means the jobs are still running.

The process ID's started were 3278, 3282, 3286, 3290, 3294 and 3299

I dont know why your ps -ef | grep "sqlplus -s" didn't show them. Next time check the individual process with something like ps -fp 3278

@chubler, is there a quick way to trap "ksh: cnt=: bad number" errors?

My code is working just fine, just a couple of places it throws this bad number error, I need to see what value is being assigned, if it is not a number.

Would help if you could show the code your running. I'd assume you are trying to do something with a variable called cnt, to test if cnt has a valid numeric try something like:

Code:

[ $cnt -gt 0 -o $cnt -le 0 ] 2> /dev/null || echo "cnt has bad numeric value of: $cnt"

@Chubler, you are right, I was trying to find the number of sql processes running and assigning that to cnt variable. Here's the code:

while read sql
do

stuff "sqlplus"

let cnt=$cnt+1

if (( $cnt == 6 )) ; then
while true
do
if (( $(ps -ef|grep "sqlplus -s"|grep -v grep|wc -l) >= 6 )); then
sleep 5
elif (( $(ps -ef|grep "sqlplus -s"|grep -v grep|wc -l) < 6 )); then
let cnt=$(ps -ef|grep "sqlplus -s"|grep -v grep|wc -l)
[ $cnt -gt 0 -o $cnt -le 0 ] 2> /dev/null | echo "cnt has bad numeric value of : $cnt"
break
fi
done
fi
done < sqlscript.lst

(In your code snippet, do we need 2 pipes? I removed one, it worked fine)

This was the output:

[1] 5970
[2] 5972
[3] 5974
[4] 5975
[5] 5978
[6] 5980
ksh: cnt=: bad number
cnt has bad numeric value of : 0
[7] 6000
[8] 6002
[9] 6004
[10] 6005
[11] 6007
[12] 6011
ksh: cnt=: bad number
cnt has bad numeric value of : 0

But our error trap should print $cnt only if $cnt -gt or -lt 0, right? Not sure why it is printing when $cnt = 0.

And it looks like the error appears when the next batch of 6 sqls are processed, starting with the second batch.