Topic: Getting random error messages on my forum

[Arantor]

You're still missing the point. You could let it run for 3 months and not see the error again. After that time would you proclaim it fixed?

The problem is that in 3 months, you might not see the error again *legitimately*. How long do you let it run for to test it? A month? Three months? Six? A year?

I have been running multiple 2.0 installations on my VPS for some years now and I have *never* seen this error on mine. I'm not denying it happens, but absence of an error is not proof of fixed, not when the error is so intermittent and sporadic.

[Vector’s Shadow]

Fine. Do nothing.

[Arantor]

What should I do, though? I can't reproduce the bug myself. I've never seen it myself. For me, doing the change would make zero difference since I cannot prove I have dealt with it!

So, I will say it again. If a user is seeing it regularly and consistently, then yes, it's worth doing to see if it makes a difference. But users who *don't* see it can't use it because there's no way to know if it actually solves the problem.

[Vector’s Shadow]

Of the listed mods, are there any that you don't have?

[Arantor]

Not relevant. There have been other cases documented where those mods are not installed (but still not consistently enough to be reproducible)

[Vector’s Shadow]

Clean installs with no mods or new themes, too?

[Arantor]

That's just it, from all the reports there is nothing in common between them other than the fact they're all errors in the caching system where a file passes the existence test and fails inclusion a statement or so later.

And it's hard to prove you fixed something when you don't have proof it's even happening to start with.


EDIT: This isn't the first time we've had such a bug. But then when it happened, for those it affected, it affected quite regularly. We could put code in to deal with it and know that sooner or later it *would* strike again. This, we can't do so.

[Vector’s Shadow]

Please tell me this happens only on servers with modsecurity.

[Arantor]

It's happened on servers with and without mod_security activated. Any other ideas? Not that mod_security would actually make any difference in this situation given what mod_security does and exactly how and where it operates.

[Kindred]

and I'll note... I have been running SMF on one or more sites for 9 years, non-stop and I have never seen this error.

[Vector’s Shadow]

Well, my general thinking right now is that you need to be running 2 instances, and in one of them you need to be on line 2658-ish while in the other just jumping into 2715. If write fails, the file would be removed by one instance and attempted to be read by the other. A rare occasion.*

Edit. *This might not be accurate. I'll check it later again when I have a little more time.

[MrPhil]

I could well believe that there is a race condition here. I documented one years ago that causes Settings.php to be emptied out (and nothing has been done about it). If a particular file is written/erased in one section of code, and another section checks for that file's existence and a statement or two later (non-atomic) tries to read it, it's certain that sooner or later an erase will slip in between the check and the read. If you've got many, many simultaneous users of a given forum, the odds become rather good that something like this will happen from time to time.

[Mikeric]

I hadn't made any changes yet, but It seems to have stopped coming up. the site is new and not busy at all. I asked a few members and I am going to keep an eye on it.

[kat]

Everything OK, still?