NT 12.45 : regular crashes on (apparently) on the same code

[AtoB]

I'm currently on NT 12.45 on several sites (used as an API-server), but my most havely used one is crashing a lot the last week. The crashing has been present since I upgraded to 12.45 (came from 9.15) last summer/fall, but not more than once every 6 weeks or so.

Currently traffic is exploding on this site (this always happens this period of the year, sometimes 5-10K requests per hour) and now it crashes 3-5 times a day and I notice that it's always the very same crash report that's generated (see attached GPFReport.txt)

I also attached a screenshot (see attached code.png) of the code (from version 12.45) it seems to point to line 1301 (but I'm not an expert in reading GPF logs ...)

I haven't upgraded to 12.55 as I could not find any fix related to this in de history docs (so to me it looks like more risks only ...)

Also problematic is that the "connectionchecker" (shipped with nettalk) I use doesn't seem to kill the exe most of the times (although it restarts a new instance ... but this new instance conflicts with the old instance).

Are there things I can do to make the connectionchecker work?

If you need additional info, let me know

[Alberto]

mhhhh
May be line 1301 must be between _Wait() and _Release() ?
I think it worth the test.
Sorry Bruce if Im wrong!

[Bruce]

Hi Ton,

>> I haven't upgraded to 12.55 as I could not find any fix related to this in de history docs (so to me it looks like more risks only ...)

Given that you're not on the current release, what exactly are you hoping I can do? Obviously your gpf.txt file is useless to me since it points to code I am not using anymore, so the line numbers are all wrong.
If you are having a problem then it's always best to be on the latest build because either;

a) the problem has already been solved, and you are struggling unnecessarily or
b) the problem still exists in which case the gpf report is useful, in conjunction with your code, and it may be possible to diagnose the cause.
c) Given that any fix made will be in the current build, not an old build, it seems like staying on the old build is untenable anyway.

Of course upgrading now when the site is especially busy is harder than doing it when you could see you had a problem, and the site was not busy. That's why it's important to do the important things (like upgrading) before they become urgent. Now you'll have to take more risk by upgrading during a busy period - but hopefully there's a lesson in there somewhere.

Cheers
Bruce