in reply to Re: simple timeout
in thread simple timeout

Well in fairness my original question wasn't why "won't alarm work", this thread has just gone that way so I've addressed the suggestions given.

It's hard to provide more detail than I already have...

The program I'm struggling in is rather big, here it is again but with the WMI server probing line included and sql queries changed for @array values to simplify. 

use strict;
use Win32::OLE qw (in);

my @servers=('servera','serverb','serverc');

foreach(@servers) {

  my $server=$_;
  $server=~s/^\s+|\s+$//g;
  my $namespace="\\root\\cimv2";

  my $object=Win32::OLE->GetObject( "winmgmts:{impersonationLevel=impe
+rsonate,(security)}//".$server.$namespace );} or die "could not get o
+bject";

  # NOW GET DRIVE INFO THROUGH WMI
}

[download]
The script goes to each server in turn, tries to initiate a WMI connection and fails on one server at the my $object line, due to some issue with the server. Instead of timing out the script just hangs there... forever. Hence the need for a script to somehow prevent this happening.

I am not asking for a solution including alarm, I've already explored using alarm to some depth, however I'm very willing to explore it further when suggestions pop up, i.e. the Time::HiRes one. I did try Win32::Process with the idea that the child could signal its parent after a set amount of time (as a substitue to alarm) but I couldn;t get the parent to seccussfully recieve a signal from it's child process.

Replies are listed 'Best First'.
Re^3: simple timeout
by BrowserUk (Patriarch) on Nov 08, 2004 at 18:56 UTC

    This is untested code. I don't have a bunch of servers against which to test.

    The basic idea is that you detch a thread that makes the connection to the server. You arrange a shared flag to be clear before the WMI attempt is made and set after.

    The main thread then waits for the flag to be set or timeout. If timeout occurs, then it just leaves the thread hanging. It will get cleaned up when the process exits. This is not particularly elegant, but iThreads lacks a defined API for killing a thread.

    I do have a mechanism that will do this by using the Win32API TerminateThread() which seems to work okay in my limited testing, but it could leave Perl in a confused state. Better I think to let any nasties happen when the process is going to exit anyway. 

    #! perl -slw
use strict;
use threads qw[ async ];
use threads::shared;
use Win32::OLE qw (in);

my @servers = qw[ servera serverb serverc ];
my $namespace="\\root\\cimv2";

for my $server ( @servers ) {
    my $gotObject : shared = 0;
    my $gotInfo   : shared = 0;
    my %info      : shared;
    async {
        my $object=Win32::OLE->GetObject( 
            'winmgmts:{impersonationLevel=impersonate,(security)}//'
             . $server
             . $namespace
        ) or die "could not get object";
        
        $gotObject = 1;

        # NOW GET DRIVE INFO THROUGH WMI
        $info{ somekey } = 'somevalue';
        $info{ someotherkey } = 'someothervalue';
        
        ## Yada yada

        undef $object;

        ## Last thing before exiting thread.
        $gotInfo = 1;
    }->detach;
    
    my $timeout = 30;
    sleep 1 while $timeout-- and not $gotObject;
    unless( $timeout ) {
        print "Failed to make connection to $server";
        next;
    }
    sleep 1 until $gotInfo;
    
    print "Received from $server";
    print "$_ => $info{ $_ }" for sort keys %info;    
}

    [download]

    The next logical progression once your using threads would be to overlap the connections to all the servers and queue the results back to the main thread for final disposal. The reason I am not suggesting this is that some OLE objects are not designed for being called from multiple threads. Also, I have experienced some problems with using Win32::OLE from multiple concurrent threads.

    Sorry it's not a tested solution. If it works, it would be most useful to those whom follow after you if you could post back a working copy of this test code with any corrections you have to make. Thanks.

    Examine what is said, not who speaks.
    "Efficiency is intelligent laziness." -David Dunham
    "Think for yourself!" - Abigail
    "Memory, processor, disk in that order on the hardware side. Algorithm, algorithm, algorithm on the code side." - tachyon
[reply]
[d/l]
Re^3: simple timeout
by dwilson@d7net (Initiate) on Nov 08, 2004 at 17:59 UTC
    How 'bout this?
    I'm making the assumption that you can create a process out of ...->GetObject(.);
    On my win2K box this creates the processes every 5 secs. you can kill them in the if stmt if necessary... 
    use Win32::Process;
...
...
my $pID;

foreach (servers) {
...
...
Win32::Process::Create ($pObj, Win32::OLE->GetObj..., ....);
$pID = $pObj->GetProcessID();
$pObj->Wait(5000) # 5 secs

// it's either been 5 secs or $pObj finished
if ($pID is there) {
    then it's been 5secs
    cleanup and restart your loop
} else {
   $pObj finished?
}
}

    [download]
[reply]
[d/l]

In Section Seekers of Perl Wisdom