I have used EDB for decades with very few issues.

My databases have got more complicated, with more relational structures, and I run queries which reference the INFORMATION and CONFIGURATION databases much more now than I used to.

Recently I have a situation where "after a while" (ie after 20 - 30 users have been using a system for a while, logging on and off, doing work etc.) New users start to see the above error #300, which completely locks them out of starting up the application. If they try 2 or more times they "get in" and it is fine. The issue seems to "get worse" (ie the first time the error occurs they will definitely get in on the second attempt, but later they may have to try 2 or 3 times).

The Delphi Code in my Apps is pretty careful about opening and closing queries and posting data so as not to leave connections hanging, but something has started to happen to bring about this error more often.

Do other users see this error, and if so what have you done to reduce it?

Any ideas for the types of action in an App that would make this problem worse?

Adam,

It is possible that as your database have grown larger and more complex, the transactions last a bit longer (more relations, index, etc) than the used to and blocking is now affecting you?

Don't know in which version you currenlty are, but seem to recall the one of latests feature Tim had added had to do with non block information and config catalogs.


Regards,
Mario

Adam


As a retiree (yippeeee) I'm F/S only these days so no recent experience, just some questions

- what does the system log events table show for around the time things are happening?
- are people logging on/off frequently or do they logon for the day?
- what's the network like?
- is there a pattern of it happening to a user / group of users? (in one job the same person could lock up 15 terminals at a time)


Roy Lambert

Roy & Mario,

Thanks for these responses. Tim has also added that it might be due to my database having multiple connections using different catalog paths. I am 100% sure it is not this, as I can replicate the problem with a single user repeatedly opening and closing the same application, after a dozen or so openings the chance of the error occurring rises.

I am using the latest version, and Mario is right that I think this was supposed to reduce blocking when accessing catalog information.

Roy: I will look at event-log data and see what I can find out. People's log on / log off behaviour varies. I think I am seeing this problem more badly now as a customer with a higher "churn" of logging on / off has upgraded to a new version, and they are the main ones complaining.

I am concerned that something is not happening when I shut down an application. (Perhaps where there is an AV or termination) I am seeing large numbers of files in the "TEMP" folder of the database (many thousands, up to 12 -15 thousand). These are all deleted by EDB in normal operation, but I see them building up in number, and not being deleted.

There is a correlation between the numbers of Temp files and the likelihood of seeing the error.

For note: In normal operation my Apps create about 500 "temp" files in the server Temp folder, but in normal operation these all delete on exit, so the issue is situations where there is a forced termination.

The customer seeing the issue is based in rural Africa and their LAN network & power is less reliable than other customers, which may be causing involuntary disconnections.

So a side issue is what can I do to get EDB to automatically clean up these TEMP files? Doing that might solve the problem.

--

** Also Roy, good to hear you have retired, but I hope you continue to hang around here a bit! You have been such a stalwart, and really saved me with EDB issues over the years. I should send you a retirement present!

Adam

Adam


I never experimented to see if the server would clean up temp files on dead session cleanup - might be worth you doing so. If not write a bit of code to delete them yourself and run as a job.

This

<<The customer seeing the issue is based in rural Africa and their LAN network & power is less reliable than other customers, which may be causing involuntary disconnections.>>

may well be the cause of the problem. What's the dead session cleanup time set to?

Roy Lambert