Server Stability

[Bruce]

There have been a couple threads lately about server stability. Obviously there are innumerable potential reasons for why a program is unstable, but I thought I'd share some solutions I've found in one place.

1) Edit AbUtil.Clw (c:\clarion8\LibSrc\win\abutil.clw)
This applies more to Windows programs than web programs, but could also affect Web programs in specific cases.
In ABUTIL.CLW is the INIClass and the TranslatorClass. Both make the error of including ASSERTs inside procedures that contain a CriticalProcedure.

the short version - comment out any ASSERT which occurs in a method that contains a CriticalProcedure.

the longer version - Asserts are a form of user interface. User interfaces should never happen inside a bit of code which is protected by a CriticalSection, or CriticalProcedure. This is especially true if you happen to Hook the ASSERT into a normal Clarion Window (or hook the MESSAGE into a normal clarion window).

I dislike editing shipping ABC files, but this has been reported to SV and will hopefully be fixed by them in future releases.

2) run as Exe
If the program is unstable as a service, then consider running it as an EXE for a period of time. Running it as an Exe serves two purposes; (a) if it's more stable that gives you some clues and (b) if it's unstable for a reason, it may show you a message telling you what that reason is.

[Devan]

This may have nothing to do with anything, but...

We had a couple of sites where the NetTalk server would GPF and shut down from time to time, for no real explicable reason, and at no particular consistent point.  One site especially would call up about twice a week with an issue.  Both sites were not 'heavy' users of the portals we created - only the occasional updates.

To try and be proactive (and reduce embarrassment to us) I decided to implement a monitoring solution which would let me know when the sites went down so that I could (hopefully) reset them before the client realised.  I used the wonderful, free, UpTimeRobot.com site for this.

The thing is - we've had UptimeRobot running for over two weeks now, and NONE of the sites have failed.  At all.

I am not sure if the 5 minute polling that UptimeRobot does tends to keep the server 'alive' and precludes the condition that causes it to crash?

Like I said - long shot, but we've certainly noticed a big improvement in stability whilst our NetTalk servers in the field get a regular HTTP hit every 5 minutes.

I'm not shilling for UptimeRobot here.  It is just the site we chose to use.  I am assuming any other monitoring site/solution that does 5 minute pings might work too.

If you have a problematic site - try setting up a monitor with either UTR or some other one, and see if that improves it... Worth a shot...

Cheers,
Devan