|
The request was a simple one. A job ran every night, moving
transaction data from an application database to a new database
that allowed users to view the source documents. The project
manager asked for email to be sent so he would know if the job
ran successfully or not.
The job consisted of two PL/SQL
procedures in a database package being run from a SQL script on
a Unix server. A log file was already being produced, but it was
not being sent to anyone and only contained the following
output:
PL/SQL procedure successfully completed.
PL/SQL procedure successfully completed.
As the developer of this
process, I knew exactly what should be in the email to be sent.
A couple of hours of development, and the log file now contained
the following (data changed to protect the innocent):
Create repository start time 02-MAY-2006
06:11:11
A01
N11
New repository count
2
Create repository end time 02-MAY-2006
06:11:12
PL/SQL procedure successfully completed.
New data start time
02-MAY-2006 06:11:12
Records read
19330
Records written
19300
Errors
6
Ignored
24
New data end time 02-MAY-2006
06:12:25
--
Update start time
02-MAY-2006 06:12:25
Records read
317
Records written
314
Errors 3
Update end time
02-MAY-2006 06:12:28
PL/SQL procedure successfully completed.
Now the log contains a lot
of information...how long the job ran, how much data was
processed, errors that might need to be researched, etc....lots
of information I proudly forwarded the new log file to the
IT manager, who approved and forwarded it to the project
manager.
And the response
was..."But all I wanted was yes or no, did it run or not!"
(Philosophical question...If the job does not run at all, how does it send
notification of said fact...but I digress. The sed / awk one
line thing is still pretty cool.)
The IT manager and
I talked about the fact that success or failure could easily be
determined from this log and that it was important to know that
errors had occurred. So we decided to have further discussions
with the project manager about the log. Half jokingly, I
suggested that I was sure I could write something really techie,
using awk or sed, to parse the file and return
either yes or no in an email. And the answer was, "You
have my approval. Have fun."
With permission to
turn my inner geek loose, I decided not only to make it happen,
but to do it all in one line, with the following result.
sed -n '/'successfully'/p'
job.log | wc -l | awk '{if($1 == 2) print "yes"; else print
"no"}' | mailx -s "Yes or No" user@yourcompany.com
This line works as follows
-
sed -n
'/'successfully'/p' job.log - sed reads the file job.log
and prints only the lines that contain the string
successfully. Since there are two PL/SQL steps in this
process, success means the string is found in two lines.
Anything else means the process did not run successfully.
The lines retrieved by sed are passed to step 2..
-
wc -l - Count the
number of lines passed by step 1. The count is passed to
step 3.
-
awk '{if($1 == 2)
print "yes"; else print "no"}' - If the count from step
2 equals 2, then print the string yes. If the
count is anything else, print the string no. This
result is passed to step 4.
-
mailx -s "Yes or No"
user@yourcompany.com -
The output from step 3, a single word yes or no,
becomes the body of an email sent to user@yourcompany.com
with the subject line Yes or No.
A challenge well met, and sed and
awk used in the same line. It was a very good day.
Note: This tip was tested using
Solaris 9, korn shell.
Was this tip useful? Did you find any errors? Do you have any suggestions? Do you care? Click
here for the tip feedback page. Thank you.
|
