Re^5: avoiding a race (the ado, you do, so well:)

[BrowserUk]

I don't see how you justify that the race doesn't result in extra e-mails. I believe your analysis is mistaken there.

Hm. The window your "still racy" referred to, was the time between a failed stat, and open. Ie:

open ERROR, '>', $file unless -e $file;
close ERROR;

And even in a full directory and an a loaded system, that time is going to be measured--assuming you can actually measure it at all--in low milliseconds at the most.

Now, what does that actually mean?

It means that one of the other processes encountered the same error as you, and succeeded in creating the error file within those few milliseconds. So what?

You then immediately overwrote it with a later time-stamp. But still, so what?

Nothing! Because the error file got created. Nothing is going to take any action--like sending emails--as a result of that files creation for another hour. From the OP:

if it has been encountered before and the time stamp is greater than an hour ago it will mail the admin

So the very worst affect of some other process creating the file instead of you, is that the sending of the email is delayed by the difference between the original time-stamp, and the new one. And that's just a few millseconds at most.

#! perl -slw
use Time::HiRes qw[ time ];
use threads qw[ stack_size 4096 ];

my $file = 'theFile';

async{ 1 until -e $file; }->detach for 1 .. 300;

sleep 3;

my @times = time;
unless( -e $file ) {
    push @times, time;
    open FILE, '>', $file or die "$file : $!";
    push @times, time;
    close FILE;
}
push @times, time;

print for @times;

unlink $file;

__END__
[20:51:05.40] C:\test>junk49
1285789900.703
1285789900.87509
1285789901.02394
1285789901.324

With 300 clients stating theFile (in a directory containing 1000 files), the window of opportunity for this irrelevant race condition is all of 300 milliseconds.

And then only if the time-stamp resolution of the file-system is sufficient to actually discern the difference, which is unlikely.

[tye]

And even in a full directory and an a loaded system, that time is going to be measured--assuming you can actually measure it at all--in low milliseconds at the most.

No, not at all. Opening the file requires finding the file which requires traversing the (possibly long) directory contents yet again (and thus contending with all of the mutex contention again also). With NTFS or a newer Linux file system (with the proper options enabled), then the directory won't be stored as a simple list and the performance is probably not as easily pathological. A few months ago I again ran into a directory with way too many files it in and it took many seconds, even minutes, to open a file (or to remove one). I haven't tried to replicate the problem on a more modern filesystem to see how well it scales. But I suspect there are plenty of file systems left in the world that were built without hash/tree directories.

And then only if the time-stamp resolution of the file-system is sufficient to actually discern the difference, which is unlikely.

And there you have your broken analysis, again. If X and Y fail to find 'file1' and then both create it, then the fact that the timestamp is not changed by whichever attempt is second has no bearing on the fact that both X and Y will then go on to send an e-mail. (Or, you can remove the race.)

[BrowserUk]

I again ran into a directory with way too many files it in

So, don't put time-critical files in huge directories.

[BrowserUk]

both X and Y will then go on to send an e-mail.

No. They won't. Because the email isn't sent until an hour later.

Are your eyes okay?

[tye]

Ah. A tiny change to the original specification leads to "notify of errors but never more than once per hour for the same error", which is something I've seen done many times and is an obvious enhancement for a notification system and how I interpreted it.

Not notifying at all unless an error repeats after at least an hour is a strange specification but that is the original proposal as written. That also leaves an open question of what the window (if any) is for when you can purge ancient error messages (which is surely not just sensible but required for the system to not eventually just fall over).

As written, the spec is also broken because (unless you add to it), once an hour has passed you get an e-mail for every single repeat of the same error (which is clearly not what is desired based on the justification given).

And that still leaves a race with multiple e-mails, just a slightly different one. If X and Y both find 'file1' with a timestamp of more than 1 hour ago, then they will both e-mail about the same error at the same time. If you remove the race, then only the one that manages to update the timestamp of 'file1' would send the e-mail (for the next hour).

And, no, I'm not convinced that my misreading of the original spec isn't actually what is really desired. (The original spec, as written, is clearly not what is really desired.)