Thread going into error condition (using MLP)

[Rob Lindsey]

I am having an issue with 4 vendors (CPSI, Soft, NetLims and Orchard) that are using a Windows system as the one making connection to our CL system.  We are on AIX 5.1 CL 3.8.1P.  The CPSI vendor says the system that they are on is going into a backup at 3 am and shuts down all apps when it does this.    The other vendors have similar situations but I have also seen when they shut down their connection to us that the interface will go into error.  To me when I turn up the EO all I see is that the CL thread tries to read the port and it is no longer there and then it goes into an error condition.  What this looks like to me is that the Windows OS is not sending a close port to our OS so that we can shut down gracefully.

Has anyone run into this before?

Thanks for any and all information.

Topic

[Rob Abbott]

A packet trace should give a good idea of what’s going on here…

iptrace on the AIX side

Ethereal (http://www.ethereal.com) on the windows side.

My guess is they’re either sending nothing or a RST or something…in any case it really should recover and not go to error state.

Can you post a log file snippet?

Rob Abbott
Cloverleaf Emeritus

[Anonymous]

Can’t be sure this applies to you, but I had a similar issue at a site several years ago. The symptoms were random engine crashes in the wee hours of the night – usually around 3AM – but not every night.

As the engine is running, it would encounter an error trying to access one of the system files (usually the db files) and would go into panic mode and shut everything down.

Eventually discovered the issue to be a file locking problem where a backup system was coming to our Cloverleaf server and backing up all the files – locking them while perform the file read.

Once the backup system changed to not exclusively access the files, the problem was solved.

Hope this helps.

Sven Williams

[Author]

Replies