Webserver stops serving at similar time each day (exe service is still running)

[Stu]

Hi Everyone/Bruce,

Got a strange one that is pretty wierd, cant work it out yet.

Since Sept 12 between 12am and 12:10am my public site webserver will stop serving (Uptime Robot pings me), although the exe service is still running when I go in to restart.

The build on Sept 11 didn't have anything dramatic in changes to the public site or dlls. And it wasn't the Nettalk 10 update (10.06) that came out on Sept 11 because I've only just upgraded to that this week.

To try and understand what is going on, I put a Debug() log into the ProcessLink embed for times >= 11:40pm and <= 12:20am.

These logs show activity right up until 12am on the dot, then one single entry for 12am and that's it. Nothing else from then until I restart the server.

Any ideas how I might carry on trouble-shooting this one?

[Bruce]

so the server is still running, you can tab from one tab to another, performance, logs and so on, but it just doesn't response to the port?

And if you click the Restart Server button does it come back?

cheers
Bruce

[Stu]

Hey Bruce,

It's usually running as a service. So I'm not sure whether that is the case or not.

I'll try tonight, leave it running open, stay up "late" and monitor.

[peterH]

Hi Stu,

Maybe a long shot, but is there something else happening on the server at that time?

Peter

[oggy]

Maybe I can help. I had similar problems with service. Occasionally, service stop working. Better say, responding. After months and months of examinations,
I realize that service stoped to response in one condition:
If service is on server, and if that service is logged as some account, for example Administrator, and if someone else, in my case myself, use that account to log on server via RDP, after logging of the RDP session, the service stop to respond. I create for that service another user account, just for service himself.
After that, never saw my service(s) stoped responding.
Edit: maybe you use one account for your service, and for some scheduled task, like backup or so....
Just my thought.
Regards, Oggy

[Stu]

Peter - Yeah maybe, will be checking the event logs. Thanks!

Oggy - Yeah I ran into that issue a few years back with NT webservers, particularly around print drivers etc, so it's running as a service with a separate account. But there might be something there, perhaps that account is having issues. Thanks!

[Stu]

Well it was 12:00 and 5,6,7 seconds when things got interesting.

I ran the public site as an exe straight (not service), and at 12:00:05 up popped a "Nettalk Object Error" (attached screenshot).

Code: [Select]

A network communication error has occurred.
Error = The URL for the Fetch command was blank.
The error number was -15 which means No servers.
Error occurred in function NetWebClient.Fetch
And there seem to be three debugview lines that are related:

Code: [Select]

[st] [netTalk][thread=2] w_Property END
[st] [netTalk][thread=1] xxx  self.state=1
[st] [netTalk][thread=1] NetSimple.ErrorTrap -15 The URL for the Fetch command was blank NetWebClient.Fetch
Now to my knowledge this doesn't happen at any other time. It certainly doesn't happen throughout the day while the public site is under the bulk of it's usual load.

So. Bruce - Does this give any further prompts to what might be going on? (remembering that this happens at the same time, once a day, each day)

[Bruce]

It's the automatic certificate check - ie the same problem you're having getting certificates at the moment.
We'll do a bit more digging on the server itself on Monday.

cheers
Bruce

update: I've also tweaked the class for 10.07 so that it does not popup the error message (unless the server is set to popup error messages.) and it also handles the error case of the GetDirectory command failing better.

[Stu]

Cool.