Reusable Server Kernel
Problem 6 (Minor): Worker API Distributed IUCV
==================== 1998-08-26 14:39:29 (-04.00) ==================== Name: ** Suppressed ** Company: ** Suppressed ** Telephone: ** Suppressed ** E-mail: ** Suppressed ** Text: I put in a new command for the WORKER command set. This command will let the operator decide whether the machines of a given class are to be managed as all-local (where the RSK does the autologging) or as potentially distributed (where the server author or admin takes responsibility for making sure they are logged on and available). For the latter case, the RSK suppresses all of its XAUTOLOG/FORCE logic and just tries the IUCV CONNECT. If it works, great, else the application gets an error. For the former case, the RSK tries to XAUTOLOG as appropriate and FORCE when necessary to reset an unresponsive worker. This setting applies on a per-class basis so that some classes can be RSK-managed while others are managed by external means. ==================== 1998-08-26 13:03:44 (-04.00) ==================== Name: ** Suppressed ** Company: ** Suppressed ** Telephone: ** Suppressed ** E-mail: ** Suppressed ** Text: It's more complicated than this, unfortunately. The RSK decides whether to XAUTOLOG based on the result of a CP QUERY USER command - if the worker machine is found not to be logged on, the RSK does the autolog and tries the IUCV CONNECT a little later. In other words, the RSK tries its IUCV CONNECT only if CP QUERY USER first reports that the userid is logged on. If your intent is to put the worker machines on another system in the ISFC collection (autologging them yourself) then the RSK will still not be able to find them because CP QUERY USER will return the status of that userid on YOUR node. I suppose I could add a configuration option that would inform the RSK to skip the CP QUERY USER check and just blindly try the IUCV CONNECT instead. I need to give this some more thought. ==================== 1998-08-26 11:31:30 (-04.00) ==================== Name: ** Suppressed ** Company: ** Suppressed ** Telephone: ** Suppressed ** E-mail: ** Suppressed ** Text: At SHARE last week someone asked whether the worker API stuff would work with worker machines located on another node in the ISFC collection (that is, would Distributed IUCV work). I commented that I didn't think I had done anything explicit to shut off Distributed IUCV but that I would check and report back. I inadvertently coded LOCAL=YES in my IUCV CONNECT for the worker API. This forced worker machines to be on the same system as the main server. I've changed this to LOCAL=NO but am not refreshing the beta. Note that if you elect to put workers on other systems in the ISFC collection, you must make sure they are autologged because the RSK will not be able to issue the XAUTOLOG command and have it take effect on another system.
[ Help | Summary | Previous | Next ]