Re: Nasty Locking-Problem magic 8.30 and PSQL8


Steven Blank
 

Pervasive Auto Reconnect (PARC) has to be enabled in the PCC on both sides, client and server.

On the server, in Configure Local Engine:

and on the client workstation, in Configure Microkernel Router:

I recommend enabling PARC on the server and leaving it that way, then enabling it only on the problematic client(s) as needed.

Unfortunately, PARC isn't always able to reconnect a client, depending on whether or not the client continues to the use same 16-byte ClientID in its calls to the Btrieve API. In other words, if Magic crashes, the next time Magic starts up, it will gen another random ClientID and the server won't be able to match the returning client with an abandoned session. Typically, if Magic reports the lost communication, it's already gone too far to successfully reconnect. PARC is only designed to Band-Aid momentary network drops and, when successful, should prevent any such errors from even appearing in Magic.

Did you enable PARC on both sides and restart Pervasive Services on the server?

Steve Blank


On 5/26/2020 9:24 AM, dm@... wrote:
Hello

A customer is using magic 8.30 application with PSQL 8 (Btrieve) and we have some nasty connection problems.
From time to time (sadly more often lately) the customer network has connection drops...  long
enough to make the application struggle. It closes with an "connection to host", fetch error or Fatal Error message.

Wouldn't be a problem if after restarting erverything would work... but we then have a deadlock... the record which
was opened stays locked and never resolves.. The only help so far was to 1. reboot the server or 2. delete the user in
the pervasive monitor.

Is there a setting or something to prevent those deadlocks? a timeout maybe? I'm tired of having my sleep stolen
to kill a user in a database monitor....

The Auto-reconnect setting doesn't seem to work or change anything...

Would be glad if someone know this issue and has a solution for it :)
or something to resolve those locked records by script?
Restarting the server is not an option while several user work with the application

Kind regards

Join main@magicu-l.groups.io to automatically receive all group messages.