Exspans Systems Inc Logo home
 
Forum
Sign up Calendar Latest Topics
 
 
 


Reply
  Author   Comment   Page 2 of 2      Prev   1   2
automan

Avatar / Picture

Moderator
Registered:
Posts: 136
Reply with quote  #16 
Question about handling x37 abends.

At the moment when the file is full or cannot be expanded anymore AutoMan puts out an informative message. It only does this once. Would it be better if it put it out for every write attempt to the dataset that fails?
0
Zamin

Avatar / Picture

Member
Registered:
Posts: 67
Reply with quote  #17 
Depending on what work is being done, you could fill the console with messages. You can easily send an email in response to a message if someone really needed to know about it and take action, so I would just put out the one message when the dataset id first filled and leave it at that.
0
rakesh

Member
Registered:
Posts: 52
Reply with quote  #18 
zOS puts X37 error messages when ever something tries to the file which is already full. And it does it every time when there a new attempt not just for the first time alone.

I feel we need to follow the same standard.

If automan is failing to write to dataset, then the admin tries to look for the messages in logs imminently. It's not normal that we go to the beginning of the log and find the error when it came for the first time.

Just my opinion.
0
automan

Avatar / Picture

Moderator
Registered:
Posts: 136
Reply with quote  #19 
rakesh, that is not strictly true. When a file is full the application abends with a system return code of x37 and the application then has the option of issuing a message when it handles the abend. The purpose of an automation program is to intervene and take action when user identified events occur, so that operator intervention is needed less often. Very few organizations have console operators anymore watching the console and responding to events. If administrators are having to scan logs it means they have more work to do in defining messages to be interpreted. As Zamin points out it is the norm for serious events that need human intervention to send an email or pager alert to a responsible person. In AutoMan's case the x37 only happens once during each allocation and that status is recorded in the file control block. That file control block remains until the file is deallocated. That means that it simply looks at the control block and if an x37 had occurred, it just does not try to write any more and so the x37 does not reoccur. The x37 will not then occur again until the file is deallocated, reallocated and a record written. So the question is, if a file is detected to have had an x37 abend, should subsequent writes display a message.
0
Grazillda

Avatar / Picture

Member
Registered:
Posts: 48
Reply with quote  #20 
Most applications crash after experiencing an x37 abend and do not try to write any more records. When the application is resubmitted it will then crash again. Obviously AutoMan cannot be allowed to abend, so it is going to recognize the abend and display a message of some sort. As it has not itself abended it will try to perform each command and so the next time the user program issues a WRITE command it is going to try to write again. If the WRITE is in a loop that could be multiple times. It should be like an application program which displays the message when the x37 happens, then stops trying to write, like an application would, then when the script is restarted display the message again when the x37 happens. So each script segment between an ALLOC and DEALLOC should be like an applications program, which crashes once when the x37 is experienced and stops writing. Then when the event that runs the script causes it to ALLOC again, it should crash again and write the message again.
0
automan

Avatar / Picture

Moderator
Registered:
Posts: 136
Reply with quote  #21 
Grazillda, this is basically how it works now. Each time there is an ALLOC-WRITE to a full dataset it display one message saying the dataset is full. As AutoMan does not crash the way an application would, it simply does not attempt to write more. But as AutoMan does not crash, it can potentially keep trying to obey a user command to WRITE, but fail the WRITE. If there are multiple intercepts on a message that does ALLOC-WRITE-DEALLOC, each one will fail on the ALLOC-WRITE and display one message.
0
automan

Avatar / Picture

Moderator
Registered:
Posts: 136
Reply with quote  #22 
The file allocation and write is now working in every section of AutoMan except SYSOUT private processing. It has to be made to work the same in all contexts, even though it is highly unlikely that a SYSOUT/private user would want to do something like this. But you never know.

When this is done I will look at all the other suggestions made for file handling. I can see doing a read next. Then maybe some of the more advanced ideas that zboxassist has suggested.
0
Previous Topic | Next Topic
Print
Reply

Quick Navigation:


Create your own forum with Website Toolbox!