[Oracle-l] OT: bash vs. ksh subprocess counts

I would be interested in seeing the output of ps -ef |grep -w $0 as well to
see what processes you see. Run it both with KSH and BASH and you may find
the difference there.
hth
Alan.-
(prevent overquoting)

[...]
I believe this is related to the fact that some shells (the KornShell
and Z-Shell) thread the last command in a pipeline differently:


[[email protected] t]$ bash -c '(ps -ef|grep -w $0)'
drado 2560 2559 0 20:17 pts/0 00:00:00 -bash
drado 3164 2560 2 20:42 pts/0 00:00:00 bash -c (ps -ef|grep -w $0)
drado 3165 3164 1 20:42 pts/0 00:00:00 bash -c (ps -ef|grep -w $0)
drado 3167 3165 3 20:42 pts/0 00:00:00 grep -w bash

[[email protected] t]$ ksh -c '(ps -ef|grep -w $0)'
drado 3168 2560 3 20:42 pts/0 00:00:00 ksh -c (ps -ef|grep -w $0)
drado 3170 3168 1 20:42 pts/0 00:00:00 grep -w ksh


From the man pages (KornShell 93 Manual Page -
www2.research.att.com/sw/download/man/man1/ksh.html):

A /pipeline / is a sequence of one or more /commands / separated by *|*.
The standard output of each command but the last is connected by a

/pipe/ <http://www2.research.att.com/sw/download/man/man2/pipe.html>(2)
to the standard input of the next command. Each command, _except
possibly the last_, is run as a separate process; the shell waits for
the last command to terminate.


P.S. I though that this difference was relative only to shell built in
commands tough ...

WOT = Wide Open Throttle

Sometimes applies to Oracle I guess.

There are better methods for doing this.

Following is a function that creates a PID file and uses it for locking.
Stale locks are handled.

a possible issues with this - not useful as is on cluster, as other
nodes could run the script.

Anyway, it may be of some use.

Jared Still
Certifiable Oracle DBA and Part Time Perl Evangelist
Oracle Blog: http://jkstill.blogspot.com
Home Page: http://jaredstill.com

==== script =====

9:49-poirot:ts20:jkstill-22 > expand -t3 locktest.sh
:

LOCKFILE=/tmp/testlock.lock

function script_lock {
typeset MY_LOCKFILE
MY_LOCKFILE=$1

# remove stale lockfile
[ -r "$MY_LOCKFILE" ] && {
PID=$(cat $MY_LOCKFILE)
ACTIVE=$(ps --no-headers -p $PID)
if [ -z "$ACTIVE" ]; then
rm -f $MY_LOCKFILE
fi
}

# set lock

if (set -o noclobber; echo "$$" > "$MY_LOCKFILE") 2> /dev/null; then
trap 'rm -f "$MY_LOCKFILE"; exit $?' INT TERM EXIT
return 0
else
echo "Failed to acquire $LOCKFILE. Held by $(cat $LOCKFILE)"
exit 1
fi
}

function script_unlock {
rm -f "$LOCKFILE"
trap - INT TERM EXIT
}

script_lock $LOCKFILE

echo press '<ENTER>...'
read dummy

script_unlock

Thx Jared - I appreciate the followup and sharing different methods!
I meant to initially share that the original solution is flawed and my intent is not to look for better methods of doing this but more for understanding details about the differences in ksh and bash with this particular situation. As others have suggested, I removed the count ("-c") from the if-test just to see better what's going on.

In bash the commands:

set -x
ps -ef|grep -w $0
if [ "`ps -ef|grep -w $0`" > "3" ]; then
exit 1
fi

... display:

+ ps -ef
+ grep -w ./bash.sh
oracle 24865 28645 0 09:33 pts/0 00:00:00 /bin/bash ./bash.sh
oracle 24869 24865 0 09:33 pts/0 00:00:00 grep -w ./bash.sh
++ ps -ef
++ grep -w ./bash.sh
+ '[' 'oracle 24865 28645 0 09:33 pts/0 00:00:00 /bin/bash ./bash.sh
oracle 24870 24865 0 09:33 pts/0 00:00:00 /bin/bash ./bash.sh
oracle 24872 24870 0 09:33 pts/0 00:00:00 grep -w ./bash.sh' ']'
+ exit 1

In ksh the same commands display:

+ ps -ef
+ grep -w ./ksh.sh
oracle 24954 28645 0 09:33 pts/0 00:00:00 /bin/ksh ./ksh.sh
oracle 24956 24954 0 09:33 pts/0 00:00:00 grep -w ./ksh.sh
+ ps -ef
+ grep -w ./ksh.sh
+ [ oracle 24954 28645 0 09:33 pts/0 00:00:00 /bin/ksh ./ksh.sh
oracle 24958 24954 0 09:33 pts/0 00:00:00 grep -w ./ksh.sh ]
+ > 3
+ exit 1

So it appears that bash forks an extra process for a command stream within an if-test while ksh doesn't. Replacing the command stream with a literal the 2 shells match up.

DAVID HERRING
DBA
Acxiom Corporation
EML [email protected]
TEL 630.944.4762
MBL 630.430.5988
1501 Opus Pl, Downers Grove, IL 60515, USA
WWW.ACXIOM.COM<http://www.acxiom.com/>


[Friend Us on Facebook]<http://www.facebook.com/acxiomcorp> [Link Us on LinkedIn] <http://www.linkedin.com/groupRegistration?gid)01735> [Follow Us on Twitter] <http://twitter.com/acxiom>

[cid:[email protected]]
________________________________
The information contained in this communication is confidential, is intended only for the use of the recipient named above, and may be legally privileged. If the reader of this message is not the intended recipient, you are hereby notified that any dissemination, distribution or copying of this communication is strictly prohibited. If you have received this communication in error, please resend this communication to the sender and delete the original message or any copy of it from your computer system. Thank you.

From: Jared Still
Sent: Friday, April 20, 2012 11:55 AM
To: Herring Dave - dherri
Cc: [email protected]
Subject: Re: OT: bash vs. ksh subprocess counts


On Sun, Apr 15, 2012 at 10:45 AM, Herring Dave - dherri wrote:
I realize this is way off topic (probably should be titled "WOT:...") for Oracle but #1, I don't currently belong to any good linux/redhat forums and #2, the issue found was from an Oracle maint. script :-)

WOT = Wide Open Throttle

Sometimes applies to Oracle I guess.

Anyway, I'm on a RHEL 4.6 server (2.6.9-67.0.22.ELlargesmp) and noticed a given DBA maint script wasn't running. It turns out that the only difference between when it last ran and now is someone changed the shell to be "bash" -> #!/bin/bash vs. #!/bin/ksh. The script has a little if-test to start with, used as a way to determine if a previous iteration of the job is still running:

if [ `ps -ef|grep -cw $0` -gt 3 ]; then
echo "$0 is already running"
exit 2
fi

There are better methods for doing this.

Following is a function that creates a PID file and uses it for locking.
Stale locks are handled.

a possible issues with this - not useful as is on cluster, as other nodes could run the script.

Anyway, it may be of some use.

Jared Still
Certifiable Oracle DBA and Part Time Perl Evangelist
Oracle Blog: http://jkstill.blogspot.com
Home Page: http://jaredstill.com
==== script =====

9:49-poirot:ts20:jkstill-22 > expand -t3 locktest.sh
:

LOCKFILE=/tmp/testlock.lock

function script_lock {
typeset MY_LOCKFILE
MY_LOCKFILE=$1

# remove stale lockfile
[ -r "$MY_LOCKFILE" ] && {
PID=$(cat $MY_LOCKFILE)
ACTIVE=$(ps --no-headers -p $PID)
if [ -z "$ACTIVE" ]; then
rm -f $MY_LOCKFILE
fi
}

# set lock

if (set -o noclobber; echo "$$" > "$MY_LOCKFILE") 2> /dev/null; then
trap 'rm -f "$MY_LOCKFILE"; exit $?' INT TERM EXIT
return 0
else
echo "Failed to acquire $LOCKFILE. Held by $(cat $LOCKFILE)"
exit 1
fi
}

function script_unlock {
rm -f "$LOCKFILE"
trap - INT TERM EXIT
}

script_lock $LOCKFILE

echo press '<ENTER>...'
read dummy

script_unlock