Network ESL

Posted by: Lee813

Network ESL - 10/03/00 07:07 AM

Currently, we're using the Network ESL.

Here's the gist of the problem - we're crossing signals something fierce at the ESL (we assume its the ESL...). If we both happen to run a problem at the same time, any or all of the following symptoms hit us. One, we get crossed into the other's file. ie, I'm working on Calc 03, he's working on 05. we run them such that the processing times overlap, and when we get to the output, more often than not, I'll be looking at the output from 05 and he'll be looking at the output from 03 (the file names will have switched in the title bar). alternatively, it'll give us the correct output, but when we return to the input screen, THEN the calcs are switched on us (or we might BOTH end up in the same file), which is considerably more insidious, because then we can inadvertently alter nodes that we think are on our calc, and completely screw up both (it's happened already, but been minor). Outside of the crossing of files, usually it follows (or just starts with, sometimes) that one or both of us come up with a fatal error, application error, Dr Watson, or combination thereof. A frequent fault is an application error in PIERCE.EXE, or something similar. Sometimes, like right now, it crashes so hard that we're both down for the count, and it can't find the ESL at all for a few minutes.

Could there be some configuration error on the server? Has this happened to any of you guys before? Please advise. Thanks.
Posted by: Richard Ay

Re: Network ESL - 10/03/00 07:20 AM

This has absolutely nothing to do with the ESL. Your problem is that you have multiple users sharing the same data directory.

CAESAR II is not one single EXE, rather it is a collection of about 45 individual programs. These programs communicate through data files. Some of these data files are job specific, such as 'calc03._a' and 'calc05._a'. Other data files are scratch and communication files used by the program, such as 'coef', 'control', and 'kdiag'.

So, when two users share the same data directory, they are fighting over these communication files. At one point, 'control' contains the job name 'calc03', but midway through its analysis, the second user updates 'control', and now it contains 'calc05' as the job name. Naturally, CAESAR II will crash - the file system is hopelessly confused.

The solution to this situation is to configure each workstation so that these communication files are unique. You can do this via the Configuration Module. From the "Tools" menu, select "Configure/Setup". On this dialog, click on the "Miscellaneous" tab. Near the bottom, find the option for USER_ID. Set the users initials here, such as "abc". (Do this for each user/workstation that will be working in this shared directory.) Exit and save the configuration.

Now, instead of using files such as 'coef', 'control', and 'kdiag', CAESAR II will use 'coef.abc', 'control.abc', and 'kdiag.abc'.

Additional posts (and examples of the configuration screens) can be found here on this Discussion Forum by following these two links:

http://www.coade.com/ubb/Forum1/HTML/000019.html
http://www.coade.com/ubb/Forum1/HTML/000020.html


------------------
Regards,
Richard Ay (COADE, Inc.)


[This message has been edited by rich_ay (edited October 03, 2000).]