Re: ADSM-L Digest - 11 May 1997 to 12 May 1997

Automatic digest processor <LISTSERV AT VM.MARIST DOT EDU>  writes:
>Date:    Mon, 12 May 1997 11:37:44 -0400
>From:    "Pittson, Timothy ,HiServ/NA" <tpittson AT HIMAIL.HCC DOT COM>
>Subject: Re: ADSM Recovery Log filling up
>
>Chuck,
>        I don't think it's entirely fixed - I'm running ADSM version 2.1.5.12
>and just experienced this problem last week...
>
>Tim Pittson
>tpittson AT himail.hcc DOT com
>
This is very interesting.  Have you contacted ADSM support on this and openned
a PMR for this?  If so could I get the PMR Number.  One of the things I did
here
to check for this 'checkpoint' problem was to write a script that would check
for this condition, (in addition check to make sure that I could get a
session).
I will include the script at the bottom incase anyone is intrested in it.  It
uses a program call gscpage to do the paging.  You can use the paging program
of your choice.

Chuck T.

#!/afs/prism/@sys/bin/perl5
#
# Description:
# This scripts checks to see if it can get an admin session with the server.
# If it can't a page is sent out to the who2page list.  It also checks the
# Check Point Log size to make sure it does not get too big.
#
# Dependencies:
# - The sysop password is pulled from the /home/root/.passwd.adsm file.
#
# USAGE: ckdsm.session.pl [<server>]
# CRONTAB Entry:
# 0,30 * * * * /home/root/bin/ckdsm.session.pl >/dev/null# Check ADSM session.
#

$threshold=500000;           # Threshold for when to send page alert.
$alarm_time=300;             # this is 5 mins.
$alarm_pid=0;                # used to set the child process id.
$Mail_to="root bob\@host.ibm.com";   # who to send mail notifications to.
$who2page_file="/home/root/who2page.txt";  # File with a list of ids to use to
                                           # for paging.

$ADMIN="sysop";              # ADSM Admin Account to use.
#Get Password for the ADSM Admin Account.
open(PWD,"/home/root/.passwd.adsm");chomp($APWD=<PWD>);close(PWD);

if ($ARGV[0]){$SERVER="-se=$ARGV[0]";print "SERVER = $SERVER","\n";}

# Check to see if the TCP/IP/shared memory interface is up by using the
# dsmadmc command.  If it fails retry $loop_limit times before returning
# with a non 0 return code ($rc).
$loop_limit=3;
$loop_count=0;
$rc=999;
while ( $loop_count < $loop_limit && $rc ne 0 ){
    &alarm_on;                      # Set an alarm to go off in 5 mins
    # Check to see if the TCP/IP/shared memory interface is up.
    $out=`/usr/bin/dsmadmc -id=$ADMIN -password=$APWD $SERVER show dbv `;
    $rc=$?;
    &alarm_off;                      # Turn the alarm off.
    if ( $rc ne 0 ){
        print "Sleeping for a while and will try again.\n";
        sleep 30;          # Wait 30 seconds and try again.....
    }
    ++$loop_count;                   # increment loop counter.
}

# Check Session access.
if ( $rc ne 0 ){
    # send page about not being able to get a session.
    $line="Unable to get Admin Session.";
    print $line,"\n";
    &sendpage;
    &sendmail;
    exit;
}

# While I have the data let's check the Checkpoints
# Get the CheckPoint Log size
if ( $out =~ /CkptLogBytes=(\w+)/ ) {$ckptlogbyte=$1;}
# Get the Interval that CkPt should happen
if ( $out =~ /CkptInterval=(\w+)/ ) {$ckptinterval=$1;}

if ($ckptlogbyte > $ckptinterval ) {
    # send mail if CkPt is above the Interval value. A CkPt should have already
    # happened or is about to happen.
    $line="Ckpt Log Byte ($ckptlogbyte) > Ckpt Interval ($ckptinterval)";
    &sendmail;
}

if ( $ckptlogbyte > $threshold && $ckptlogbyte > $ckptinterval ){
    # If the CkPt Log is bigger then the threshold we have big trouble. PAGE
    # someone and let them know.
    $line="Database Checkpoint Log is too big: $ckptlogbyte";
    &sendpage;  # Send out a page of failure.
}

print "CkptLogBytes = $ckptlogbyte \nCkptInterval = $ckptinterval \n";
exit;
# End of main prg.


sub sendpage {
    # send pages to list of people in the who2page_file.  The text of the
    # page is passed in with the $line variable.

    open(PAGE, $who2page_file);   # open the who2pag file and read in the list
    while ( <PAGE> ){             # of gsc ids and push them on a stack.
        chomp;
        push(@who2page, $_);
    }
    close(PAGE);

    # page each person on the stack, poping off that id until the stack is
    # empty.
    while ( $i=pop(@who2page) ){
        # Send out a page of failure.
        `gscpage $i \"$hostname $line\"`;
    }
    print $line,"\n";

    print $out;
}

sub sendmail {
    # send mail to let someone know what happened.
    # $Mail_to, $line, $out are passed in.
    open(MAIL,"|sendmail -t -f adsm-admin ");
    print MAIL "To: $Mail_to\n";
    print MAIL "From: adsm-admin\n";
    print MAIL "Reply-To: nobody\@austin.ibm.com\n";
    print MAIL "Subject: $HOST: $line \n\n";
    print MAIL $out,"\n";
    close(MAIL);

}

sub alarm_on {
    # the idea here is to turn the alarm on, then run the command.  If the
    # alarm is not turned off after the command is ran then the command
    # never finished or it "timed out".  A page will be sent to say as much.

    # create a child process to be the alarm
    if (!( $child = fork) ) {
        print "Alarm set : $child\n";
        # here in the child process the work is done.
        sleep $alarm_time;
        $line="dsmadmc command timed out";
        &sendpage ;
        exit;
    } # end of alarm process.
    $alarm_pid=$child;
}

sub alarm_off {
    # go kill the child process before the alarm goes off.
    kill 9, $alarm_pid;
    print "Alarm truned off : $alarm_pid\n";
}
# End of perl script.