I open the file extend I have been tasked with finding data there are certain 
cases that are not easily found so I have to search a large number of files on 
the off chance that 2 jobs write to the duke if it’s corrupted I can run it 
again it won’t happen and like  I said it’s not production if it was that would 
change things 


> On Mar 26, 2021, at 9:54 AM, Paul Gilmartin 
> <[email protected]> wrote:
> 
> On Fri, 26 Mar 2021 09:21:59 -0400, Joseph Reichman  wrote:
> 
>> The chances of two jobs writing to the file are slim and none I am searching
>> a large number of files because I cannot find the data
>> 
> https://en.wikipedia.org/wiki/Murphy%27s_law
> 
>> And most importantly it is NOT a production job
>> 
> It seems unusual to be "searching a large number of files" in what
> "is NOT a production job".
> 
> What is your procedure for detecting contention?  What is your recovery 
> process?
> 
> I'll suggest using for output a UNIX file opened with O_APPEND.  Be sure
> your writes are atomic.
> 
> 
>> -----Original Message-----
>> From: Jeremy Nicoll
>> Sent: Friday, March 26, 2021 9:13 AM
>>   ...
>> That won't work if two jobs simultaneously try to write to the file.
>> 
>> You might not destroy the file, but you will lose the information that the
>> second job tried to write to it.
>> 
>> The jobs need
>> 
>> (a) not to allocate the output dataset at all in the JCL
>> 
>> (b) to have a retry loop that does something like the
>>   following when you want to write data out
>> 
>>  repeat
>>      try to get exclusive use of the results file
>> 
> If you have authority, use S99WTDSN  #  ##########################
> 
>>      if that fails and we've tried many times,
>>         perhaps write the data somewhere else?
>>          - though that might also fail
>>         stop trying
>>      end
>> 
>>      if exclusive use isn't acquired but we've not tried
>>         often, wait a bit (a random time interval) and
>>         try again
>> 
>>      if exclusive use was granted, update the file,
>>         free it again, and stop trying
>> 
> Else?  # ##################################
> 
>>   end
> 
> -- gil
> 
> ----------------------------------------------------------------------
> For IBM-MAIN subscribe / signoff / archive access instructions,
> send email to [email protected] with the message: INFO IBM-MAIN

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to